[squid-users] SQUID reverse-proxy configuration for large Web apps

Hi,

I need some advises to design SQUID as a reverse proxy for many web
apps(~ 20 servers)

There will be 500 users using this system 24/7.
Each users will have 2 permanent connections using chunk encoding and
1 connection using HTTP keep alive mechanism.

Finally Squid will have to process 1000 simultaneous permanent
connections and 500 "control" connections.

An other point is that squid will have to rewrite the HTTP Location
header to support HTTP 302 redirection.

I made some tests with the Apache ab product on a 2.6 STABLE 6.

I tests 1000 simultaneous connections using keep alive. each
connections request 3 times the same url. I aims to test SQUID in a
heavy load and simulate 1000 permanent connections

This test, unfortunatly, was not really good: 2838 requests failed.

Concurrency Level: 1000
Time taken for tests: 44.672 seconds
Complete requests: 3000
Failed requests: 2838

I were 3 main errors during the test :
X-Squid-Error: ERR_SOCKET_FAILURE 24
X-Squid-Error: ERR_CONNECT_FAIL 71
X-Squid-Error: ERR_CANNOT_FORWARD 0

Is there a limitation on SQUID for simultaneous users like limited tcp
port range or something else ?

thanks

Sylvain
Received on Fri Jul 18 2008 - 14:38:48 MDT