Closed tantaman closed 1 year ago
Looks like this improved write TPS by ~15%
The reason is that we no longer need to consult the database at the start of each write transaction as we actively update the cached db_version. If there are two writers each taking turns then we'd lose the perf improvement here. So the perf improvement 15% in single writer scenarios and between 0 and 15% in multi-writer scenarios.
…version after commit
This also allows us to stop clearing the db version cache and should give us quicker TPS.