Re: What exactly is postgres doing during INSERT/UPDATE ?

On Mon, Aug 31, 2009 at 10:31 AM, Kevin
Grittner<Kevin(dot)Grittner(at)wicourts(dot)gov> wrote:
> Jeff Janes <jeff(dot)janes(at)gmail(dot)com> wrote:
>> Joseph S <jks(at)selectacast(dot)net> wrote:
>
>>> The question is what I do with my 14 drives. Should I use only 1
>>> pair for indexes or should I use 4 drives?  The wal logs are
>>> already slated for an SSD.
>
>> Why not just spread all your index data over 14 spindles, and do the
>> same with your table data?
>
> If you have the luxury of being able to test more than one
> configuration with something resembling your actual workload, I would
> strongly recommend including this as one of your configurations.
> Spreading everything over the larger number of spindles might well
> out-perform your most carefully hand-crafted tuning of object
> placement on smaller spindle sets.

The first thing I'd test would be if having a separate mirror set for
pg_xlog helps. If you have a high write environment moving pg_xlog
off of the main data set can help a lot.