max-cache-size (was: no subject)

[JINMEI Tatuya / 神明達哉]

At Thu, 26 Jun 2008 10:53:19 +0200,
Matus UHLAR - fantomas <uhlar at fantomas.sk> wrote:

> > I'm not sure what you're indicating by this, but it's the expected
> > effect if you specify a finite max-cache-size, whether it's 9.4 or
> > 9.5.  Even if you don't specify it, that's also possible depending on
> > the query pattern, again, whether it's 9.4 or 9.5.
> 
> I was mostly curious if I can expect 9.5 eat more or less of memory (is more
> memory-effective) if I would not set finite max-cache-size under the same
> conditions.

It's difficult to answer because that would highly depend on the query
pattern and the TTLs of cached entries.  As such, this can be a
totally meaningless guess, but if everything except the max-cache-size
is configured to be the default, and if queries and the cached TTLs
are not so biased (I know this is a big 'if'), 9.5 would east less
memory than 9.4.  This is because 9.5 purges expired entries
regularly, while 9.4 performs cache cleaning only every 60 minutes.

Again, it highly depends on your specific operational environment, and
I suspect no one can provide a general answer that necessarily matches
your case.  If you'd really like to know it, the only feasible option
would be to do some experiments with actual queries you're receiving.

---
JINMEI, Tatuya