Re: redirector in squid-1.1b4

From: John Saunders <John.Saunders@dont-contact.us>
Date: Tue, 1 Oct 1996 12:40:44 +1000 (EST)

On Tue, 1 Oct 1996, Russell Street wrote:

> Having the redirectors used round-robin would also be good, otherwise
> you get:
>
> redirector #1: 136584 ## (consumed 3m 48s cpu time)
> redirector #2: 5867
> redirector #3: 1567
> redirector #4: 352
> redirector #5: 73
> redirector #6: 24
> redirector #7: 11
> redirector #8: 5
> redirector #9: 4
> redirector #10: 2

I'm not sure that round robin would gain anything. Since the 1st
redirector gets the majority of requests, it has a much higher chance of
not being swapped out on a loaded system that uses swap. Also squid should
only select a redirector if it's not busy, so selected one further down
the list over the first available one doesn't gain anything.

The only reason I can think of for round robin is if the redirector
suffers bit rot after a lot of requests. Then distributing the load
equally will delay the time when it will crash. But in an ideal situation
it wouldn't matter if a redirector has handled 1000000 or 1 request, it
should work equally well.

Cheers.
-- +--------------------------------------------------------------+
        . | John Saunders - John.Saunders@scitec.com.au (Work) |
    ,--_|\ | - john@nlc.net.au (Home) |
   / Oz \ | - http://www.nlc.net.au/~john/ |
   \_,--\_/ | SCITEC LIMITED - Phone +61 2 9428 9563 - Fax +61 2 9428 9933 |
         v | "Alcatraz triathalon: dig, swim, run" |
              +--------------------------------------------------------------+
Received on Mon Sep 30 1996 - 19:45:27 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:33:06 MST