The connection code contains a special feature "turbo mode", which lets a worker thread continuously read a connection, without passing
it back to the polling mechanism. This can enhance performance of very active connections.
It is applied to replication connections by default, but if single replication operations like adding entries are relatively long, it could be better to apply them in parallel (the supplier already sends them asynchronously).
Investigate if applying repl ops on a consumer in parallel, can be done without breaking replication (update of ruv?, urp?) and what the
performance impact is.
Probably making turbo mode configurable helps in investigation and to adopt to specific load scenarios
allow turbo mode to be disabled
Author: Rich Megginson email@example.com
Date: Wed Feb 13 07:18:27 2013 -0700
Metadata Update from @rmeggins:
- Issue assigned to rmeggins
- Issue set to the milestone: 1.3.1
to comment on this ticket.