On 9/12/05, Christopher Petrilli <petrilli(at)gmail(dot)com> wrote:
> 2) Tune ext3. The default configuration wrecks high-write situations.
> Look into data=writeback for mounting, turning off atime (I hope
> you've done this already) updates, and also modifying the scheduler to
> the elevator model. This is poorly documented in Linux (like just
> about everything), but it's crtical.
I'm using noatime and data=writeback already. I changed my scheduler from
the default anticipatory to deadline and saw an improvement, but I haven't
yet delved into playing with specific elevator tunable values per-device.
3) Use 8.1 and strongly look at Bizgres. The data partitioning is critical.
I've just started down the path of getting 8.1 running with a larger block
size, and tommorow I'm going to look at Bizgres's partitioning as opposed to
some manual schemes. Will the current Bizgres have a lot of the performance
enhancements of 8.1 already (or will 8.1 or 8.2 eventually get Bizgres's
In response to
pgsql-performance by date
|Next:||From: Ron Peacetree||Date: 2005-09-13 05:02:16|
|Subject: Re: Performance considerations for very heavy INSERT|
|Previous:||From: Brandon Black||Date: 2005-09-13 03:44:35|
|Subject: Re: Performance considerations for very heavy INSERT traffic|