Re: [squid-users] squid rproxy - heavy load test.. summarization..

From: Henrik Nordström <hno@dont-contact.us>
Date: Tue, 13 Nov 2001 13:33:47 +0100

On Tuesday 13 November 2001 11.29, Klavs Klavsen wrote:
>
> 1) Squid simply died at approx. 900 requests..
> - is there any way I can make squid - drop connections above this
> limit - or something like that - so that some - say the first 900
> requests/sec gets served..

Squid tries to back off when given too much, but this area needs more tuning.

Note: too aggressive "tuning" of number of filedescriptors or I/O threads
defeats the backof algoriths the way they are currently designed as Squid
then does not notice the overload until it is too late.

> but if there are requested more than it can handle it simply drops
> them.. or perhaps nicely queue them - so that it doesn't go above a certain
> response time?

Implementing an backoff algorithm based on response time and request rate
would probably be a good idea.

> - I'm gonna set the AIO proc amount to 100.. does that seem fair,
> for a box that needs to deliver this kind of req/sec?

Still definite overkill for such a small amount of drives as you have.

> - Anything else I can do.. to increase the max req/sec, the max
> concurrent sessions, and the way it deals with Max loads?

There is plenty that can be done. Most are outside what I can provide as part
of the free Squid-users support.

Regards
Henrik Nordström
MARA Systems AB
Sweden
Received on Tue Nov 13 2001 - 05:33:24 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:04:10 MST