[Gllug] wwwoffle or squid? (or something else?)
Nix
nix at esperi.demon.co.uk
Wed Sep 19 23:15:52 UTC 2001
On Wed, 19 Sep 2001, John Edwards stated:
> On Wed, Sep 19, 2001 at 08:20:52PM +0100, Stig Brautaset wrote:
>> In my house we are three people that connects to Internet via a
>> dedicated machine running debian potato. The only services running is
>> currently diald (and obviosly sshd ;). This is a pentium 75 with 24
>> megabytes of ram and about a gigabyte disk free. I think it is about
>> time for it to start pulling it's own weight around here, so I want to
>> set up an http proxy/cache and wanted to know whether anybody has any
>> opinions.
>>
>> I have heard rumours about squid being the best of the best, but I was
>> thinking that maybe wwwoffle is better in my case since it is targeted
>> specifically towards dialup systems.
>
> Squid has a reputation as a resource hog, but you can adjust the RAM
> usage in the conf file. wwwoffle has some nice features such as the
> ability to use the files in the cache when the internet connection is
> down, and has a web interface for configuration. I used to use it on
> a home network but I found it occasionally gave errors connecting to
> some slow sites (eg xoom.com), so switched to squid.
wwwoffle's timeouts *are* configurable, y'know :)
wwwoffle's largest advantage as I see it is that if you find a bug in it
it tends to get fixed in minutes or hours; Andrew Bishop (the author) is
very helpful (as well as being a semi-lurker on uk.comp.os.linux).
The caching and on/offlinedom handling is nice, and it's acquiring lots
of ancillary features too, like recolouring links to pages in the
wwwoffle cache, compressing pages as they stream from web sites to
wwwoffle, from wwwoffle to the browser, and in the cache, and spotting
web-bugs ;)
(I don't know how many of these features squid has got. I don't use it.)
wwwoffle's downside is that it's meant for smaller sites so if you have
a huge cache you'll be in semi-untried territory; its storage mechanism
(`/var/spool/wwwoffle/{protocol}/{hostname}/{hash-key}') is bad for
linear-time lookup filesystems like ext2, so putting a big cache on
reiserfs (or xfs) instead is probably a good idea. I've found a couple
of bugs myself (and they've both been fixed really fast, although
supplying patches is appreciated :) )
> And installing some more RAM would be very helpful, as it can avoid
> having to swap out processes disk which will drag the performance
> down something rotten.
I dunno about that. On this (1Gb) cache on a 32Mb box wwwoffle is:
RSS VSZ COMMAND
892 2052 /usr/sbin/wwwoffled -c /var/spool/wwwoffle/wwwoffle.conf -d 4
so hardly a massive resource hog. Just about anyone can afford 1Mb for a
daemon.
(However that box is headless and doing little else, so...)
--
`Upsetting this BOFH was a BAD MOVE.' --- Chris Newport
--
Gllug mailing list - Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug
More information about the GLLUG
mailing list