Re: [SQU] Log Analyziers

Kareem Mahgoub wrote:
>
> Hi all,
> I have 2 very simple questions if you could help me on any of them.
>
> 1) I need a log analyzer for my squid 2.3 stable1 running on Solaris 7
> All what I need is Miss : Hit percentage plus the response time of the
> cache plus memory utilized and CPU.

http://www.squid-cache.org/Scripts/

> 2) How I can Optimize the squid.conf to get best results for the response
> time. I have the following config
> a) Sun Enterprise 450 dual processor 233 MHz
> b) Solaris 7
> c) 768 Mbytes RAM
> d) Squid 2.3 stable 1
> e) I don't run any other application on my server.
> f) I am serving 400 Clients concurrently on average.

Tricky one. Depends quite a lot on your situation. For a start, is it
the link that is the main bottleneck, or are you running into resource
limitations in the Squid server?

squid.conf directives you might want to look into:

* refresh_pattern to increase the hit ratio where there are opportunity
to do so
* quick_abort_* settings
* half_closed_clients. My recommendation is off unless you need to
support such clients (none of the browsers or HTTP command line tools
requires this)

If you are running into resource limitations in the Squid server, you
first needs to tell if those limitations are caused CPU or disk I/O.

If CPU, then the only choice is to get a faster CPU. Squid is
single-threaded and can only use one CPU, so it might be an idea to run
two Squids on the box to be able to utilize both CPUs.

If disk, then look into async-io or diskd, and make sure you have
sufficient number of harddrives to manage the I/O load. You might also
get an improvement from using the logging function of disksuite.

--
Henrik Nordstrom
Squid hacker
--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html