[database] Support multiple (non-best) chains, concurrent reads on different forks. #255
Labels
Comments
|
I don't fully understand the issue here but would it mean that |
|
In short, the ask here is for the database to support storage of orphan blocks/side chains. Currently, it only ever stores blocks which are on the main chain. Changing this will require a pretty significant amount of changes in other packages because they all work off the documented assumption that asking the database for blocks and transactions are only available if they have already been fully validated and are in the current best (main) chain. The description here is extremely misleading because it's referring to btcd when it should be referring to RPC. That said, we're ultimately going to want to allow storage of the side chain blocks as it will allow a few more advanced techniques to be used for optimizations and allow RPC access to information about reorged chains which is really what this issue is asking for. |
|
Ok thanks for the explanation I understand now. I completely agree that storing side chains is a good aim. Being able to detect and monitor side chains as they develop on the Bitcoin network also has potential security benefits for clients of btcd who want to limit their exposure to chain reorgs. A consequence of storing side chains might make it possible for the |
|
As an update to this, the current plan is to start tackling this within the next few weeks. We have been discussing several of the design implications and the current plan is as follows: The btcdb interface is going to change quite significantly. The current interface and approach has a few undesirable properties such as the following:
To resolve the above issues (and others not covered), the current plan is to redesign the interface so it provides better scalability, more easily facilitates new backends such as BigTables for GAE, and opens the door for supporting more advanced features such as block pruning. To accomplish this, the idea is to have the new interface provide two major categories as follows:
Finally, these changes mean that most of the current Bitcoin-specific intelligence, like spend tracking, ensuring the block being inserted extends the main chain, and unspending transactions when blocks are dropped, will no longer live in btcdb and instead will be moved to a higher layer where they belong. |
|
Overall, I'm very pleased with @davecgh's proposed changes to Some of the design flaws/quirks with the current interface have been made rather apparent as I've been taking a more critical look at IMO the new separation of concerns (metadata vs blocks) that the proposed changes would bring about are much needed. The old interface appears to have been designed primarily with LevelDB in mind, as well as storage of metadata and blocks stuffed into the same location (which forces all drivers to have I look forward toward to the ultimate interface re-write whatever shape it may take. A more flexible interface a may attract new contributors seeking to add alternative databases for |
|
It was great to read @davecgh's Here's the type of architecture that, from my experience, is compatible with concurrent block chain reads and NoSQL datastores: There are a couple of points to make with the above interface:
The following is the functionality that should be calculated at a higher level by something like
Judicious use of caching for the above three functions could be possible so long as the cache is wiped as soon as Anyway, I am not sure this solution exactly addresses @davecgh's proposal however I am looking forward to see how it all evolves. |
davecgh
changed the title
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Apr 16, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary of some of the reasons for the major redesign are as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intellegience such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. - The vast majority of database activity after the initial block download is read activity, however leveldb, as its name implies, is optimized for levelled write performance as opposed to read performance. This is really the exact opposite of the desired characteristics for a bitcoin database. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich meatdata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of differente mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffboltdb (flat file plus boltdb metadata backend). Here is a quick overview: - Highly optimized for read performance - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - The metadata storage uses boltdb under the hood which provides fast memory-mapped access - NOTE: Due to the fact the database can get quite large and due to how the OS does memory-mapped files this can result in the process showing huge amount of memory used. This is typical for memory-mapped files, but the key thing to remember is the reported memory is not _really_ used because the OS will relinquish it as soon as other processes need it - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Apr 16, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary of some of the reasons for the major redesign are as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intellegience such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. - The vast majority of database activity after the initial block download is read activity, however leveldb, as its name implies, is optimized for leveled write performance as opposed to read performance. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffboltdb (flat file plus boltdb metadata backend). Here is a quick overview: - Highly optimized for read performance - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - The metadata storage uses boltdb under the hood which provides fast memory-mapped access - NOTE: Due to the fact the database can get quite large and due to how the OS does memory-mapped files this can result in the process showing huge amount of memory used. This is typical for memory-mapped files, but the key thing to remember is the reported memory is not _really_ used because the OS will relinquish it as soon as other processes need it - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
4 tasks
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Apr 16, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary of some of the reasons for the major redesign are as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intellegience such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. - The vast majority of database activity after the initial block download is read activity, however leveldb, as its name implies, is optimized for leveled write performance as opposed to read performance. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffboltdb (flat file plus boltdb metadata backend). Here is a quick overview: - Highly optimized for read performance - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - The metadata storage uses boltdb under the hood which provides fast memory-mapped access - NOTE: Due to the fact the database can get quite large and due to how the OS does memory-mapped files this can result in the process showing huge amount of memory used. This is typical for memory-mapped files, but the key thing to remember is the reported memory is not _really_ used because the OS will relinquish it as soon as other processes need it - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Apr 17, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. - The vast majority of database activity after the initial block download is read activity, however leveldb, as its name implies, is optimized for leveled write performance at the expense of read performance. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffboltdb (flat file plus boltdb metadata backend). Here is a quick overview: - Highly optimized for read performance - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - The metadata storage uses boltdb under the hood which provides fast memory-mapped access - NOTE: Due to the fact the database can get quite large and due to how the OS does memory-mapped files this can result in the process showing a huge amount of memory used. This is typical for memory-mapped files, but the key thing to remember is the reported memory is not really used because the OS will relinquish it as soon as other processes need it - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Apr 17, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. - The vast majority of database activity after the initial block download is read activity, however leveldb, as its name implies, is optimized for leveled write performance at the expense of read performance. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffboltdb (flat file plus boltdb metadata backend). Here is a quick overview: - Highly optimized for read performance - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - The metadata storage uses boltdb under the hood which provides fast memory-mapped access - NOTE: Due to the fact the database can get quite large and due to how the OS does memory-mapped files this can result in the process showing a huge amount of memory used. This is typical for memory-mapped files, but the key thing to remember is the reported memory is not really used because the OS will relinquish it as soon as other processes need it - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Apr 30, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. - The vast majority of database activity after the initial block download is read activity, however leveldb, as its name implies, is optimized for leveled write performance at the expense of read performance. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffboltdb (flat file plus boltdb metadata backend). Here is a quick overview: - Highly optimized for read performance - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - The metadata storage uses boltdb under the hood which provides fast memory-mapped access - NOTE: Due to the fact the database can get quite large and due to how the OS does memory-mapped files this can result in the process showing a huge amount of memory used. This is typical for memory-mapped files, but the key thing to remember is the reported memory is not really used because the OS will relinquish it as soon as other processes need it - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
May 1, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. - The vast majority of database activity after the initial block download is read activity, however leveldb, as its name implies, is optimized for leveled write performance at the expense of read performance. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffboltdb (flat file plus boltdb metadata backend). Here is a quick overview: - Highly optimized for read performance - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - The metadata storage uses boltdb under the hood which provides fast memory-mapped access - NOTE: Due to the fact the database can get quite large and due to how the OS does memory-mapped files this can result in the process showing a huge amount of memory used. This is typical for memory-mapped files, but the key thing to remember is the reported memory is not really used because the OS will relinquish it as soon as other processes need it - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 7, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. - The vast majority of database activity after the initial block download is read activity, however leveldb, as its name implies, is optimized for leveled write performance at the expense of read performance. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffboltdb (flat file plus boltdb metadata backend). Here is a quick overview: - Highly optimized for read performance - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - The metadata storage uses boltdb under the hood which provides fast memory-mapped access - NOTE: Due to the fact the database can get quite large and due to how the OS does memory-mapped files this can result in the process showing a huge amount of memory used. This is typical for memory-mapped files, but the key thing to remember is the reported memory is not really used because the OS will relinquish it as soon as other processes need it - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 17, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 17, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 17, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 17, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 17, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 28, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 28, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 28, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jul 28, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Aug 26, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Aug 26, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Sep 6, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Sep 17, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Oct 16, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Oct 16, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Oct 16, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Oct 16, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Oct 23, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Oct 24, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Oct 26, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Nov 14, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Nov 28, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Dec 8, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Dec 10, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Dec 29, 2015
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Jan 6, 2016
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Feb 1, 2016
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Feb 1, 2016
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Feb 3, 2016
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Mar 16, 2016
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future.
davecgh
added a commit
to davecgh/btcd
that referenced
this issue
Mar 16, 2016
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future.
3 tasks
TheSusort
pushed a commit
to qchain/btcd
that referenced
this issue
May 17, 2016
This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in btcsuite#255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes btcsuite#255 - Fixes btcsuite#199 - Fixes btcsuite#201 - Implements and closes btcsuite#256 - Obsoletes and closes btcsuite#257 - Closes btcsuite#247 once the required chain and btcd modifications are in place to make use of this new code
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Thursday Jul 17, 2014 at 04:34 GMT
Originally opened as btcsuite/btcdb#6
The database currently only saves the best chain. If a fork extends beyond the current best chain, the reorged blocks will be removed from the database. This makes the database unsafe for multiple callers who concurrently expect their given block range to be valid.
For example, when btcd must rescan some range of blocks, and the tip of the range is removed from the best chain due to a new best block, db lookups will begin failing for missing the block for some hash, or worse (if only heights are used), lookups will silently continue with no indication that previously processed blocks may not be on the same fork.
A better solution would be to use an iterator which allows walking through some range of blocks in a fork. Even if the best chain is switched, these blocks can still be iterated over. After the iteration is finished, the caller can finish (it returned results about the chain at the time the request was made) or a switch back to the main chain and/or best block can be made if more blocks were attached.
The text was updated successfully, but these errors were encountered: