Re: [squid-users] cache dir equation + Misc questions?

From: Brian <hiryuu@dont-contact.us>
Date: Mon, 15 Apr 2002 01:24:26 -0400

cachme_mem is entirely separate from the disk index. It holds hot and in
transit objects. You will want to keep it large enough to handle the
active load at peak time, but not much more in a memory situation this
tight.

I assume you actually meant 256 in the cache_dir line you posted?

There are lots of possibilities here.

Are you swapping (run 'vmstat 2' for a bit to see what's swapped in and
out)? How big is the actual squid process? What CPU are you on?

You may want to test aufs, instead of ufs. Squid is probably blocking on
disk IO a LOT, since the OS can't effectively cache 60GB in ~200MB of
cache.

I'm not sure how quick the hashing is in squid but, with around 4.5
million objects, memory latency is bound to catch up with you eventually.

        -- Brian

On Monday 15 April 2002 12:51 am, Anthony Giggins wrote:
> OK in that Case if I have 60GB of cache_dir it should need 60 * 10 =
> 600MB RAM.
>
> The server its self has 1GIG RAM so I don't see a problem here but do I
> need to change anything in the config, for example
>
> cache_mem do I set this to 600?
>
> Thanks for your help
>
> -----Original Message-----
> From: Hermann Strassner [mailto:hermann.strassner@hama.de]
> Sent: Friday, 12 April 2002 6:45 PM
> To: Anthony Giggins
> Subject: RE: [squid-users] cache dir equation + Misc questions?
>
> Some more:
> Squid has an index of the files in cache_dir in memory. It uses about 76
> Bytes per file, and the average size of the objects in Internet is about
> 13 kB (11.5 kb here with us).
> So it needs about 6 MB of memory per 1 GB cache_dir plus some extra
> space only for the index.
> Calculate 10 MB of memory per 1 GB of cache_dir.
>
> Hermann
> -----Original Message-----
> From: Anthony Giggins [mailto:AGiggins@synergyit.com.au]
> Sent: Friday, April 12, 2002 6:00 AM
> To: squid-users@squid-cache.org
> Subject: [squid-users] cache dir equation + Misc questions?
> Is there a cache_dir equation to work out the most efficient use of 1st
> and 2nd level directories? I'm running a large cache ie. 60GB and I
> find that my first and second level Directories fill-up and as a result
> proxy access slows down. The cache volume is no where near full. Ie
> about 13GB after about 100Days Usage and we average between 0.5 - 1 GB
> per day.
>
> cache_dir ufs /cache 60000 128 156
>
> most other options are just the defaults
>
> I'm running squid 2.3.STABLE4 on redhat Linnux 7.1
>
> Any help would be great.
>
> Anthony Giggins
Received on Sun Apr 14 2002 - 23:24:28 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:07:33 MST