Re: Now I am back, next thing. Final PGS tuning. - Mailing list pgsql-general

From John R Pierce
Subject Re: Now I am back, next thing. Final PGS tuning.
Date
Msg-id 49DCD719.1000509@hogranch.com
Whole thread Raw
In response to Re: Now I am back, next thing. Final PGS tuning.  (Jennifer Trey <jennifer.trey@gmail.com>)
List pgsql-general
Jennifer Trey wrote:
> Scott, thank you.
>
> I think I might have misunderstood the effective cache size. Its
> measured in 8kB blocks. So the old number 449697 equals 3.5 GB, which
> is quite much. Should I lower this? I had plans to use 2.75GB max. Can
> I put 2.75GB there? Should I leave it?

effective_cache_size is an estimate of how much disk data the OS is
likely to have cached in memory.   postgres uses this to guess whether
or not recently read data is likely to be 'fast' (in the system cache)
or 'slow' (on the physical disk, hence requiring disk IO to read).
This value is used in some fairly abstract heuristics, it does NOT need
to be that accurate, its jusr a ballpark estimate.

you should run your system under your expected workload, then view the
actual working cache size in Task Manager ("System Cache" on the
Performance tab of the task manager in XP, I dunno about 2008
Server)...  Now some of that cache probably belongs to other processes
than postgres, so round down a bit.   On my desktop system at the
moment, I'm showing 1.3GB



pgsql-general by date:

Previous
From: Radcon Entec
Date:
Subject: Re: Table has 22 million records, but backup doesn't see them
Next
From: Radcon Entec
Date:
Subject: Re: Table has 22 million records, but backup doesn't see them