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
