wal_buffers, redux - Mailing list pgsql-hackers

From Robert Haas
Subject wal_buffers, redux
Date
Msg-id CA+TgmobgMv_aaakLoasBt=5iYfi=kdcOUz0X9TdYe0c2SZ=2Pg@mail.gmail.com
Whole thread Raw
Responses Re: wal_buffers, redux
Re: wal_buffers, redux
List pgsql-hackers
I've finally been able to run some more tests of the effect of
adjusting wal_buffers to values higher than 16MB.  I ran the test on
the 16 core (x 4 hw threads/core) IBM POWER7 machine, with my usual
configuration settings:

shared_buffers = 8GB
maintenance_work_mem = 1GB
synchronous_commit = off
checkpoint_segments = 300
checkpoint_timeout = 15min
checkpoint_completion_target = 0.9
wal_writer_delay = 20ms

I ran three 30-minute tests at scale factor 300 with wal_buffers set
at various values from 16MB up to 160MB, in multiples of 16MB, using
pgbench with 32 clients and 32 threads in each case.  The short
version is that 32MB seems to be significantly better than 16MB, by
about 1000 tps, and after that it gets murky; full results are below.

16MB tps = 13069.420658 (including connections establishing)
16MB tps = 14370.293228 (including connections establishing)
16MB tps = 13787.219680 (including connections establishing)
32MB tps = 14916.068968 (including connections establishing)
32MB tps = 14746.448728 (including connections establishing)
32MB tps = 15110.676467 (including connections establishing)
48MB tps = 15111.981870 (including connections establishing)
48MB tps = 12824.628192 (including connections establishing)
48MB tps = 15090.280081 (including connections establishing)
64MB tps = 15382.740815 (including connections establishing)
64MB tps = 12367.942312 (including connections establishing)
64MB tps = 15195.405382 (including connections establishing)
80MB tps = 15346.080326 (including connections establishing)
80MB tps = 12791.192216 (including connections establishing)
80MB tps = 14780.804054 (including connections establishing)
96MB tps = 15476.229392 (including connections establishing)
96MB tps = 15426.012162 (including connections establishing)
96MB tps = 15548.671849 (including connections establishing)
112MB tps = 15400.669675 (including connections establishing)
112MB tps = 15676.416378 (including connections establishing)
112MB tps = 15016.651083 (including connections establishing)
128MB tps = 15811.463522 (including connections establishing)
128MB tps = 15590.343669 (including connections establishing)
128MB tps = 15256.867665 (including connections establishing)
144MB tps = 15842.131696 (including connections establishing)
144MB tps = 15669.880640 (including connections establishing)
144MB tps = 15753.460908 (including connections establishing)
160MB tps = 15658.726637 (including connections establishing)
160MB tps = 15599.600752 (including connections establishing)
160MB tps = 15311.198480 (including connections establishing)

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company


pgsql-hackers by date:

Previous
From: Robert Haas
Date:
Subject: Re: pg_prewarm
Next
From: Tom Lane
Date:
Subject: Re: pg_crypto failures with llvm on OSX