Re: Hardware performance for large updates

Josh Berkus wrote:
> The problem: The update series (done as a function) takes 10-15
> minutes. During this time, the CPU is never more than 31% busy, only
> 256mb of 512 is in use, and the disk channel is only 25% - 50%
> saturated. As such, is seems like we could run things faster.
>
> What does everybody suggest tweaking?
>

I think we'd need more information to be of any help -- schema,
functions, explain output, etc.

I do think you probably could increase Shared Buffers, as 256 is pretty
small. There's been a lot of debate over the best setting. The usual
guidance is start at 25% of physical RAM (16384 == 128MB if you have
512MB RAM), then tweak to optimize performance for your application and
hardware. You might also bump sort mem up a bit (maybe to 2048). Again,
I would test using my app and hardware to get the best value. Are you on
a Linux server -- if so I found that fdatasync works better than (the
default) fsync for wal_sync_method.

HTH,

Joe