To add some context to this change. If you look at the current [benchmark results in the CI](https://concourse.apachegeode-ci.info/teams/main/pipelines/apache-develop-main?groups=complete) you will notice the get based benchmarks. Those are currently pegged at this throughput rate regardless of the number of threads thrown at the client above about 32 threads. By removing the lock contention in the connection pool the throughput with ~600 threads is nearly 3 times as fast and can continue to scale up with threads and processors.
You can run these benchmarks yourself. See http://github.com/apache/geode-benchmarks. Keep in mind your mileage may vary depending on your hardware. The CI uses dedicated AWS instances with 36 processing units and the 10Gb+ enhanced networking in the same placement group. [ Full content available at: https://github.com/apache/geode/pull/3304 ] This message was relayed via gitbox.apache.org for [email protected]
