Re: performance measuring

From: Alex Rousskov <rousskov@dont-contact.us>
Date: Tue, 9 Nov 1999 08:36:04 -0700

On Tue, 9 Nov 1999, [iso-8859-9] Erdinç Dallı wrote:

> I want to measure the performance of different configurations of squid. How
> will be the performance if we use different cache sharing protocols like
> ICP, Cache digests, Summary Cache and so on. I can do this by changing the
> configuration and do the measuring. But if it is possible I want to do the
> measuring with absolutely the same request pattern. If it is not, it would
> be nice to use a resuest pattern which is taken from real operation. So is
> there a tool or method that takes squid logs as input and feeds squid
> accordingly.

Just two quick notes. If you want the same request pattern, I assume that
is because you want the workload to be the same for all of your tests.
Well, the latter is not, strictly speaking, possible if you are going to
use life servers on the Internet. Second, if you are OK with synthetic
workloads, you may give Web Polygraph a shot. You will have to build your
own workload, but it may be doable. Also, note that there is at least one
existing study comparing ICP with Cache Digests (see FAQ).

Alex.
Received on Tue Nov 09 1999 - 08:49:36 MST

This archive was generated by hypermail pre-2.1.9 : Wed Apr 09 2008 - 11:57:32 MDT