24.2 RC1 Performance Report #2176
-
|
Not an issue perse, but I wanted to provide the anecdote and offer to gather any additional debugging info that may be helpful as I recall reading that performance should be improved in this release. I just got a new piece of hardware to test that out, and have found that initial sync is still quite bad in the latest RC (downloaded binaries). In it's first 48 hours, this new system has so far only sync'd to block 650k and now validating roughly 1.25 blocks/second
addrindex and txindex are enabled CPU utilization oscillates between 15-25% with btcd using 34 threads. RAM utilization is negligible. Internet bandwidth for headers is not a bottleneck. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
(converted to a discussion)
I think you'll find that dropping the addrindex will give you a dramatic speed up. It's the largest index we maintain on disk, and will cause a lot of extra I/O that the new utxo set cache doesn't help with. |
Beta Was this translation helpful? Give feedback.
-
|
Wow, yea from the jump a 5-10x even with the nvme drive Was under the impression for some reason this was an optimal flag for neutrino clients? That not/no longer the case? |
Beta Was this translation helpful? Give feedback.
(converted to a discussion)
I think you'll find that dropping the addrindex will give you a dramatic speed up. It's the largest index we maintain on disk, and will cause a lot of extra I/O that the new utxo set cache doesn't help with.