RE: [squid-users] hyperthreading?

Dumb SMP question (not related to hyperthreading): provided that the current
2.4 Linux scheduler respects affinity generally (a process tends to stay on
same logical processor):

Let's suppose that I have a farm of DP boxes that I run Squid on as an
accelerator, and these would be balanced on another box routing incoming
HTTP connections with something like ldirectord or a
hardware-content-switch, would I get more performance if I run 2 Squid
instances on the same Physical box, each bound to a different interface, and
each using non-competing I/O & disk resources, and each having plenty of
RAM? Is this even possible? If so, desirable?

Sean

-----Original Message-----
From: Henrik Nordstrom [mailto:hno@squid-cache.org]
Sent: Thursday, January 16, 2003 2:04 AM
To: Stephen Amadei
Cc: Wei Keong; squid-users@squid-cache.org
Subject: Re: [squid-users] hyperthreading?

tor 2003-01-16 klockan 04.38 skrev Stephen Amadei:

> I won't speculate on the merits of SMP with Squid, as I have no
> benchmarks to compare.

All benchmars seen so far show the expected: Close to none, and in some
cases even negative.

Regards
Henrik
Received on Thu Jan 16 2003 - 14:57:55 MST