--MimeMultipartBoundary
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable
Oskar Pearson wrote:
> On a similar subject:
>=20
> Would anyone object if the values for cache_swap_low and cache_swap_hig=
h
> were changed to be either percentage values or actual kb values?
>=20
> This way with a 10 gig cache with a 1% difference in cache_swap_low
> and cache_swap_high you wouldn't delete 10megs at a time... given that
> some people have 40gig caches....
Yes, the difference between high-low water marks should probably be
settable directly in size and not % of cache size.. this seems like a
good idea.
And the purge method should probably contignously purge objects (when
time permits) as long as the cache is above the low-water mark (and
purge objects at full speed when over high-water). I had some thoughts
on this earlier, and the design that I thought about was to periodically
build a list of candidates for purging, and then pick objects from this
list as needed.
We should probably add some sort of min_free support as well, to make
sure that the disks never get completely filled. This requires yet
another external child process, but besides that is is almost trivial
for most platforms..
--- Henrik Nordstr=F6m --MimeMultipartBoundary--Received on Tue Jul 29 2003 - 13:15:44 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:11:29 MST