Greetengs!
Found that simple test pgbench -c20 -T20 -j8 gives approximately
for REL_15_STABLE at 5143f76: 336+-1 TPS
and
for REL_16_STABLE at 4ac7635f: 324+-1 TPS
The performance drop is approximately 3,5% while the corrected standard deviation is only 0.3%.
See the raw_data.txt attached.
How do you think, is there any cause for concern here?
And is it worth spending time bisecting for the commit where this degradation may have occurred?
Would be glad for any comments and concerns.
With the best regards,
--
Anton A. Melnikov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company