Re: [SQU] high throughput?

Squid in it's current form can handle upwards of 180 reqs/sec on nicely
sized Intel hardware running Linux and an async compile of Squid (this
much I've confirmed with benchmarks). You can probably push it to
250-300 on really really big Intel hardware--like 2GB RAM, dual 1GHz
CPU, 4-6 SCSI 15K RPM disks, etc. You'll then need to cluster
several Squid boxen together using an L4 switch, WCCP, or LVS, to
balance between the multiple caches.

The Squid developers, Henrik and Adrian in particular, are working hard
on making Squid faster and much more scalable. The 2.5DEVEL tree (open
now, as the 2.4 tree moves rapidly toward STABLE) will feature a number
of new development ideas that it is hoped will increase it's scalability
significantly. It's unlikely we will see 2.5 reach stability in 2001,
however, since it is just getting rolling.

Since you speak of changing the code, I assume you have some programming
experience...if so, please do feel free to help with the effort to speed
up Squid. I'm sure everyone would be happy to have an extra set of eyes
and hands working on accelerating the code. Check out what's going on
over at the squid.sourceforge.net site, get a feel for the code, and get
involved.

Hope this answers your questions.

dualcat@163.net wrote:

> Dear friends, I like squid.From the benchmark results,it seems that
> the max peak throughput of squid is near 140 req/sec with x86
> systems.But,in my application,I need a cache system which peak
> throughput reach 1000 req/sec.Can I do that with squid? with modify
> the code or change the OS or change the hareware? Thanks for your
> help!
>
> Jingchang Huang 1.10

                                   --
                      Joe Cooper <joe@swelltech.com>
                  Affordable Web Caching Proxy Appliances
                         http://www.swelltech.com

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