Re: about the performance of squid

From: ShuXin Zheng <zhengshuxin@dont-contact.us>
Date: Tue, 30 Jan 2007 14:34:18 +0800

I want to test the new HEAD version very much. But I must add the
codes I've added for my special application to it, which will spend
some time. Some functions below may be useful for web-accelerate:
1) support local domain resolve by add configure item in squid.conf
such as below:
acl MY_DOMAIN dstdomain www.squid-cache.org
local 192.168.0.1 192.168.0.2:80 192.168.0.3:8080 ...MY_DOMAIN
In this configure, squid will select one of these IP of
"192.168.0.1 192.168.0.2:80 192.168.0.3:8080 " for one client request.
When one of these IP's server is error, squid can easily use the other
one and at the same time invoke one probe request to the bad error
server util it can be used again.
2) support HTTP-POST retry. This may be difficult to do because I saw
many comment in squid's codes saying that POST will not be restarted.
But in reality, post retry is necessary for the backend application web
servers.
3) improve the acl match efficiency. In a web-accelerate application, there
are many acl policies be configured in squid.conf, so one url request will
be matched with many acl lines. I think that may reult in resources wasting.
I hear the PCRE(perl regex lib) is very efficient which has self-study function,
and was be applied in POSTFIX, APACHE. Squid may add some compiling
options for regex match with PCRE.

Thanks

On 1/30/07, Adrian Chadd <adrian@creative.net.au> wrote:
> On Tue, Jan 30, 2007, ShuXin Zheng wrote:
>
> > Because I use squid in product environment and have made some changed in
> > squid, I can't test the "HEAD squid" version you suggest. Would you please
> > tell me the priority of it ? Thanks.
>
> Henrik and I are trying to keep Squid-2-HEAD as stable as possible whilst still
> allowing for development.
>
> I'm successfully running Squid-2-HEAD on a pair of somewhat busy squid servers
> at Wikimedia without any known issues.
>
> I understand that you can't run non-"stable" versions of Squid in production
> but if noone tells us the codebase has issues or problems then it becomes "Stable".
> Then you run it, and have problems :) So I'd really appreciate it if you
> could test out Squid-2-HEAD on one server or internally and provide us with
> some feedback.
>
> I'm planning on bringing over many more performance and resource related changes
> in the upcoming months which should make Squid a heck of a lot faster. They, however,
> requires -some- testing. :)
>
>
>
> Adrian
>
>

-- 
zsxxsz
Received on Mon Jan 29 2007 - 23:34:25 MST

This archive was generated by hypermail pre-2.1.9 : Thu Feb 01 2007 - 12:00:02 MST