[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: shameless advice request

On Tue, Feb 03, 1998 at 04:43:02PM +0100, Michael Alan Dorman wrote:
> One important thing to try and do is balance your RAM usage against
> your available HD space---my squid process runs about 48MB for a 2GB
> cache.  That relationship should, I think, scale linearly, since I
> think the data it holds is a fixed size, and directly proportional to
> the number of objects in the cache.
...and run the novm version, as in linux the buffer cache makes in-memory
cacheing redundant. Utilising the "lots of FDs" patch is a must, and if you
have a real big intranet than recompile squid to really use that many FDs.
(I have 10000+ client, and if the firewall gets rebooted, squid runs out of
the compiled 1024 FD limit within a minute.)
> I would generally echo Tim, although I think anything above a P5 is a
> waste of money---put that into more RAM and faster HD access.
True. I have 128M RAM, which is fine, but not a conservative setup at all,
but one of my processors have no work.
Given that it is for a 1Mbps line fully utilized, and a 10000+ company,
for your ~1000 client 128M and one processor seems to be conservative

> I've also got some code for putting hit/miss ratios to accesses and
> bytes transferred into MRTG that I should probably package up some
> day.
It would be warmly welcomed!

GNU GPL: csak tiszta forrásból

TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-devel-request@lists.debian.org . 
Trouble?  e-mail to templin@bucknell.edu .

Reply to: