Re: How to simulate heavy load?

In article <5m8dk7$75m$1@zeus.ven.ra.rockwell.com> you write:

>So, since we are far away of getting the 70000 hits/hour that cause
>problems, I was wondering if there is a way of simulating such a heavy
>load so I can see with my own eyes how Linux and Squid perform.
>
>We have a pretty fast HP-UX box so I was thinking about telneting
>to the proxy port as fast as I can a retrieve a cached document. It is
>just that I do not know how to automate the telnet. I was thinking
>about something like this in bash:

Retrieving cached documents as fast as you can doesn't say much. With a
little bit tuning you can get hundred of hits per second.

What you really want to know is how Squid behaves when your Internet
connection gets saturated or parts of the Internet are not reachable for
you. In such cases the average duration of a connection gets longer leading
to a very large amount of parallel connections to Squid.

>telnet proxy 8080 << EOF
>GET http://xxx.xxxx.xxx/ HTTP/1.0
>EOF
>
>but this does not work.

There should be an extra empty line after the GET line. See the scripts
directory of the Squid distribution for some test programs. It is very
important that you simulate a large number of parallel clients.

>Any ideas on how to put Squid and my OS to sweat?

Take a number of old Squid logfiles and send the URL's in each of them to
your Squid server using a program like tcp-banger.pl. Do not use one
logfile for all tcp-banger programs but use different ones for each
tcp-banger program.

Arjan
Received on Sun May 25 1997 - 08:58:01 MDT