On 21 sep 2009, at 23.41, Bruce Momjian <bruce(at)momjian(dot)us> wrote:
> Alan McKay wrote:
>>>> And if so, where does that extra load go? ? ?Disk? ?CPU? ?RAM?
>>> As of 8.4.X the load isn't measurable.
>> Thanks Bruce. What about 8.3 since that is our current production
> Same. All statsistics settings that are enabled by default have
> near-zero overhead. Is there a specific setting you are thinking of?
That's not true at all.
If you have many relations in your cluster that have at some point
been touched, the starts collector can create a *significant* load on
the I/o system. I've come across several cases where the only choice
was to disable the collector completely, even given all the drawbacks
8.4 makes this *a lot* better with two new features. One enabled by
default (write stats file on demand) and one you have to enable
manually (stats file location). Using both these together can pretty
much get rid of the issue, but there's no way in 8.3.
In response to
pgsql-performance by date
|Next:||From: Matthew Wakeling||Date: 2009-09-22 10:58:27|
|Subject: Re: query memory consumption|
|Previous:||From: Ron Mayer||Date: 2009-09-22 05:12:30|
|Subject: Re: Slow select times on select with xpath|