Re: [squid-users] fourth cache off??

----- Original Message -----
From: "Joe Cooper" <joe@swelltech.com>
To: "fooler" <fooler@skyinet.net>
Cc: "khiz code" <khizcode@yahoo.com>; <squid-users@squid-cache.org>
Sent: Wednesday, December 19, 2001 1:32 PM
Subject: Re: [squid-users] fourth cache off??

> > i think the main reason why squid's request per second is always around
from
> > 100 to 200 no matter how high performance hardware you are going to used
is
> > due to the squid's algorithm. my suggestion is that, its time to
re-design
> > squid. :->
>
> Great! When do you plan to start on the rewrite, fooler? ;-)

got some prototypes here :-> based on jonathan lemon's (freebsd guy)
kqueue() event instead of using select() or poll() but portability to
non-BSD OS is the main issue here :-> but i dont know when to finish this
due to other reasons because this is not my priority list. probably adrian
or duanne can code that to squid... this url can give you more info about
this kqueue event: http://people.freebsd.org/~jlemon/kqueue.pdf .. it is
worth mentioning too that as much as possible is to have a non-blocking i/o
in all aspect like disk i/o, network i/o , and others.

> j/k. Squid is probably due for an overhaul...

i must agree to this...

> And the Squid-3 discussion
> has popped up a few times on the dev list--well worth reading. There's
> even a design idea 'whiteboard' somewhere on the Squid site. I'll be
> interested to see the final design that the guys come up with.
>
> Don't forget that another reason the performance isn't increasing as
> much as raw hardware performance might indicate is that Alex keeps
> making the Polygraph workloads harder!

hehehe let us blame alex for this! :-> oh btw, merry christmas and a happy
new year to all!

fooler.
Received on Tue Dec 18 2001 - 23:17:23 MST