Saturday I was in my favorite grocery store, standing in the line, browsing the net on my phone. I read Vadim Tkachenko‘s blog post about Measuring Percona Server Docker CPU/network overhead and his findings were opposite than mine – he didn’t found any measurable difference. Reading his post, he did found huge impact in networking which I didn’t check, so I was re-run my checks again, but now with paying attention to network configuration.

I won’t linger this more: he’s correct, I was wrong. After I start to use the host’s network directly – the performance degradation went away.

Yay, for Vladim!

OLTP test statistics:
 queries performed:
 read: 140182
 write: 40052
 other: 20026
 total: 200260
 transactions: 10013 (1939.54 per sec.)
 read/write requests: 180234 (34911.72 per sec.)
 other operations: 20026 (3879.08 per sec.)
 ignored errors: 0 (0.00 per sec.)
 reconnects: 0 (0.00 per sec.)

General statistics:
 total time: 5.1626s
 total number of events: 10013
 total time taken by event execution: 41.2134s
 response time:
 min: 2.96ms
 avg: 4.12ms
 max: 38.07ms
 approx. 95 percentile: 5.21ms

Threads fairness:
 events (avg/stddev): 1251.6250/41.54
 execution time (avg/stddev): 5.1517/0.00