On Thu, 28 May 1998, Henrik Nordstrom wrote:
> Squid 1.1.X has a quite bad throughtput graph when put under high load
> due to blocking on disk I/O. Squid 1.2 improves this a great deal by
> using asyncronous disk I/O.
>
> It would be interesting to see a similar comparisation using Squid 1.2
> (with async-io enabled).
It would be interesting to see a similar comparison using Squid 1.2 with
async-io disabled as well :) Both to see how much persistent connections help
(compared with Squid 1.1) and what is the gain from using async-io (compared
with Squid 1.2 with async-io enabled).
Of course, any good performance analysis requires running a lot of tests
under different conditions/scenarios. Even if vendors do that internally,
they are interested in posting one result showing their product wins the
most. Clearly, that one "lucky combination" is virtually guaranteed to be
very rare in practice..
Alex.
Received on Thu May 28 1998 - 08:57:38 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:40:29 MST