possible effective_io_concurrency performance regression

-hackers,

While updating my Postgres performance curriculum I was doing some
testing with effective_io_concurrency and I may have found a regression.
I am aware that the parameter only works under certain conditions.
However, what I appear to have found is that if it is set to anything
but 0, it is a regression for (at least benchmarksql tpc-c) workloads.

See here:

Testing with the TPC style benchmark shows that on local systems,
setting this between 0 – 48 keeps the TPS within noise level. However,
testing also shows that on cloud systems such as Google Cloud Compute
setting this setting to anything greater than 0 results in an
approximately 10% performance degradation on TPS:

Local/GCE effective_io_concurrency TPS

GCE OFF 47951

8 43098

1 43233
LOCAL
0 9939

4 9960

16 9955

48 9958

I was able to produce these results pretty consistently. I wonder if any
has seen this on EBS? GCE instance is 16CPU, 59GB memory, 240MB
Sustained rate SSD with 15k IOPS.

Thanks,

JD

--
Command Prompt, Inc. || http://the.postgres.company/ || @cmdpromptinc

PostgreSQL Centered full stack support, consulting and development.
Advocate: @amplifypostgres || Learn: https://pgconf.us
***** Unless otherwise stated, opinions are my own. *****

--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers