Replicated Concurrency Control and Recovery

Implemented a distributed database, complete with serializable snapshot isolation, replication, and failure recovery. It gives a deep insight into the design of a distributed system.

Overview

This project involves implementing a distributed database with the following features:

• Serializable Snapshot Isolation (SSI)
• Replication using the Available Copies approach
• Failure recovery

Algorithms Used

• Available Copies approach to replication with SSI and validation at commit time
• Aborting writes on failed sites

Concurrency Handling

• Detect concurrency-induced abortion conditions due to first committer wins and consecutive RW edges in a cycle
• No need to restart aborted transactions; handled by the application