Re: Slowness of Squid without any apparent reason

On Friday 18 October 96, at 15 h 55, the keyboard of Stephane Bortzmeyer
<bortzmeyer@pasteur.fr> wrote:
                        
> Our largest cache in France is an AlphaServer 1000 (Digital Unix 4.0)
> running Squid 1.0.18. It is incredibly slow, taking several seconds to
> deliver a very short HTML page (already in the cache) and sometimes

This has not yet improved :-( I just noticed that, during night or even
lunch, the cache has a normal speed (less than a second of delay). So,
this seems related to its load but also to the actual occupation (I
cleared the cache_dir and started from zero and the cache had a normal
speed until it reached 70-80 % of the cache_dir).

> Configuration :
>
> cache_mem 130

Decreased to 110 Mb. ps shows Squid is still too large (I discovered that
"info" in cachemgr shows a lot of useful info about memory consumption.
Squid eats a lot of memory so do not use a cache_mem too close from your
physical memory, you need memory for other things). Swapping is still low.

> cache_swap 14500
>
> (We have 320 Mbytes of RAM. The cache disk is 100 % full.)

I decreased TTL and now the cache is no longer full. cache.log shows no
more garbage collecting.

David Luyer <luyer@ucs.uwa.edu.au> suggested not to use ADVFS (we don't)
for the cache and to have several separate cache-dir (we have near 7300
files in each directory which is quite slow to process) or to go to Squid
1.1 with its two-dir level. (Not yet tried.)

Redfern Ian <RedfernI@logica.com> explained that Squid is very slow when
garbage-collecting files on disk. That's why I reduced TTL but the
problem persists.

jrg@demon.net (James R Grinter) suspected swapping, noticing the process
is quite large (I wasn't even aware that in-transit objects were in
memory, there are 80 Mb of them!). Swapping is very small now but I'll
decrease again cache_mem.

Thanks to all those who searched, still looking for advices,

Stephane Bortzmeyer Institut Pasteur
bortzmeyer@pasteur.fr Service d'Informatique Scientifique
                              Paris, France
+33 01 40 61 34 62

http://web.cnam.fr/personnes/bortzmeyer/home_page.dom
Received on Wed Oct 30 1996 - 06:16:55 MST