Обсуждение: PostgreSQL 9.3.2 Performance tuning for 32 GB server
Hi All,
I have configured the blow parameters for a 32 GB server . I this correct ?
shared_buffers = 6GB
work_mem = 24MB maintenance_work_mem = 250MB
effective_cache_size = 16GB
shared_preload_libraries = 'pg_stat_statements' pg_stat_statements.max = 10000 pg_stat_statements.track = all
wal_buffers = 8MB
checkpoint_segments = 32
checkpoint_completion_target = 0.9
--
--Regards
RAMAKRISHNAN KANDASAMY
--Regards
RAMAKRISHNAN KANDASAMY
On Sat, Jan 25, 2014 at 12:02:59PM +0530, RAMAKRISHNAN KANDASAMY wrote: > Hi All, > > I have configured the blow parameters for a 32 GB server . I this correct ? > > shared_buffers = 6GB going over 2GB probably doesn't help > work_mem = 24MB maintenance_work_mem = 250MB work_mem depends a lot of your queries and the number of clients, but with 32GB RAM setting a default work_mem of 128MB would probably not hurt. Your maintenance_work_mem is too low, raise it to 2GB. > effective_cache_size = 16GB if it's a dedicated server you can raise it to 24GB > shared_preload_libraries = 'pg_stat_statements' pg_stat_statements.max = > 10000 pg_stat_statements.track = all > wal_buffers = 8MB > checkpoint_segments = 32 depends of your load, 10's reasonable for light loads. 50 or 100 isn't uncommon for heavier ones. Keep in mind that every increase of 30 will cost you 1 gigabyte of disk space in pg_xlog and an extra ~2-5 minutes (depends of your i/o) of recovery time after a crash. > checkpoint_completion_target = 0.9 > > It's considered as a bad habit to change the cost settings, but I often raise the default cpu_tuple_cost to 0.08 (instead of 0.01) too. > -- > --Regards > RAMAKRISHNAN KANDASAMY -- No trees were killed in the creation of this message. However, many electrons were terribly inconvenienced.
On Fri, Jan 31, 2014 at 8:55 AM, Julien Cigar <jcigar@ulb.ac.be> wrote:
On Sat, Jan 25, 2014 at 12:02:59PM +0530, RAMAKRISHNAN KANDASAMY wrote:going over 2GB probably doesn't help
> Hi All,
>
> I have configured the blow parameters for a 32 GB server . I this correct ?
>
> shared_buffers = 6GB
That is true on 32 bit system. On a 64 bit system with 32GB of RAM, there is a lot of value to be potentially gained by having shared buffers significantly higher than 2GB.
It's considered as a bad habit to change the cost settings, but I often
raise the default cpu_tuple_cost to 0.08 (instead of 0.01) too.
> --
> --Regards
> RAMAKRISHNAN KANDASAMY
--
No trees were killed in the creation of this message.
However, many electrons were terribly inconvenienced.
--
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance
Thomas John