RE: [SQU] redirector spec

Are you suggesting a round-robin approach? Squid currently does this:

1st request - first free redirector
2nd request - first free redirector
...

When a redirector is currently handling a request it is marked Busy.
If all the redirectors are Busy, the request is queued.

Why use first-free instead of round-robin?

First-free uses the first helper the most, then the 2nd and so on, so
most of your 32 helpers can sit swapped out freeing RAM for your working
set. Using a round-robin approach will page all your helpers into
memory. From a CPU point of view the first free is better as well
because when a request is finished, if there is a queue, squid can just
send the next request to the helper immediately, without going around a
poll() loop...

Rob

> -----Original Message-----
> From: Dimitris 'sehh' Michelinakis [mailto:sehh@altered.com]
> Sent: Wednesday, 8 November 2000 9:23 PM
> To: squid-users mailing list
> Subject: [SQU] redirector spec
>
>
> I'm running Squid 2.3.STABLE4 which runs 32 processes of the
> squirm redirector.
>
> I've noticed on high traffic that Squid doesn't use all the
> redirector processes at
> the same time. It just uses the first available, and when
> that fills up, it uses the
> second available.
>
> I think that this 'procedure' is not effective. CPU usage is
> higher when a single
> redirector process is used to its maximum, than splitting the
> work into
> all available processes.
>
> Is it possible to notifty the developers of Squid? I'd like
> to know if this change
> is possible.
>
> þ H.I.C. & D.B.S. þ OS/2 Warp þ Hellas þ
> þ ServerConfig þ ConfigEdit þ OS/2 UK UG þ
>
> --
> To unsubscribe, see http://www.squid-cache.org/mailing-lists.html
>
>

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html