Re: CLOG contention

On Dec 20, 2011, at 11:29 PM, Tom Lane wrote:
> Robert Haas <robertmhaas(at)gmail(dot)com> writes:
>> So, what do we do about this? The obvious answer is "increase
>> NUM_CLOG_BUFFERS", and I'm not sure that's a bad idea.
>
> As you say, that's likely to hurt people running in small shared
> memory. I too have thought about merging the SLRU areas into the main
> shared buffer arena, and likewise have concluded that it is likely to
> be way more painful than it's worth. What I think might be an
> appropriate compromise is something similar to what we did for
> autotuning wal_buffers: use a fixed percentage of shared_buffers, with
> some minimum and maximum limits to ensure sanity. But picking the
> appropriate percentage would take a bit of research.

ISTM that this is based more on number of CPUs rather than total memory, no? Likewise, things like the number of shared buffer partitions would be highly dependent on the number of CPUs.

So perhaps we should either probe the number of CPUs on a box, or have a GUC to tell us how many there are...
--
Jim C. Nasby, Database Architect jim(at)nasby(dot)net
512.569.9461 (cell) http://jim.nasby.net