Re: [squid-users] how many hits can squid can handle

I'm afraid Squid doesn't scale quite that well. I just shipped a unit
specced as follows:

Dual PIII 1GHz
1.5GB RAM
3 x 18GB 15k RPM disks

With an extremely fast server class motherboard from Tyan. I used a
simple IP division for load balancing between two Squid processes.

The server is running up to 240 reqs/sec without any problems, and good
performance. But CPU hits 0% idle about half the time during these high
load periods. I expect it will handle another 30-40 reqs/sec sustained
before it begins to suffer on performance.

Also note that this same hardware when operating as a single process
(plus disk io threads) can only maintain about 190 reqs/sec before
slowing to unnacceptable hit latencies (I consider anything greater than
0.04 seconds as being an unacceptable hit latency).

So I'd say even a 1.4GHz Athlon would max at 250 reqs/sec. Though a
dual Athlon 1.4 might run up to 350 or more.

Chemolli Francesco (USI) wrote:

>>4000 req/min or 4000req/sec
>>4000req/sec is well nigh impossible for squid .. but Henrick
>>knows best
>>
>
> On good hardware (say a 1.4Ghz Athlon) with enough RAM (lots of it)
> you could get as high as 450 hits/second, and 2-3 Mbyte/s (client-side)
>
> This is based on my real-life experience, where on a P3/500 (1 Gb RAM)
> I get 180 reqs/s and 1 Mbyte/s.

                                   --
                      Joe Cooper <joe@swelltech.com>
                  Affordable Web Caching Proxy Appliances
                         http://www.swelltech.com
Received on Wed Sep 12 2001 - 17:19:43 MDT