A few days ago I came across the announcement that MaxScale 1.3 finally supports Persistent Connection.
ProxySQL has supported Persistent Connection since it was a prototype (sometime back in 2013). I am therefore very happy that the MaxScale Team finally introduced a long awaited feature.
However, MaxScale's implementation has a serious drawback and I consider it a serious bug (more details towards the end of the article). A bug so serious that it shouldn't reach any GA release.
I like running benchmarks, and given the new feature in MaxScale, I thought it a good time to compare ProxySQL vs MaxScale again. It's been around 6 months since the previous benchmark.
ProxySQL vs MaxScale Benchmark setup
The benchmark setup is very simple:
- a physical server (20 CPU cores) running sysbench, proxysql and maxscale
- a physical server (20 CPU cores) running 3 mysqld …