[rescue] SS2 memory?

[Greg A. Woods]

[ On Thursday, March 7, 2002 at 20:32:24 (-0500), Joshua D Boyd wrote: ]
> Subject: Re: [rescue] SS2 memory?
>
> Hmm.  We are talking maybe 5 users on a 640k downlink DSL line.  Perhaps it
> will be totally awefull.  It seems to me that just caching images on common
> sites should be a pretty big help though.  But then, for that, a simple
> proxy written in python should do the job.

You need to look at the actual content you commonly view before you'll
be able to judge the effectiveness of an HTTP cache for that few number
of users.  There's still an awful lot of what should be static content
that's not cache-able.  Stupid, ignorant, (and lazy) web designers still
rule the Internet.

Of course the best way to gather your stats is to run an HTTP proxy and
analyse its logs.....  :-)

An SS2 as a mostly pure HTTP proxy with very little attempt to cache
anything but the most recent traffic, should work OK in your situation
(though it may indeed increase latency more than you wish to tolerate),
but you won't be able to get very far in terms of building a significant
cache.  A 32GB cache makes the squid process over 500MB.  You might be
OK with a 1G cache at most, assuming you have 64MB of RAM, though if I
were you I'd go with about 256MB at most and try to squeeze as much into
the in-core buffers as possible.  The SS2 will be fast enough to be
mostly invisible when it's getting hits out of memory.  It'll take quite
a lot of T.L.C. and analysis to get it tuned optimally though.

While there's nothing like a good experiment such as this to learn the
ropes of such technology, you'd probably be better off from a browsing
performance P.O.V. just using your ISP's cache, assuming they have one
and that it's not overloaded.  That much bandwidth will make a bigger
and more widely shared cache very effective -- i.e. it really will speed
up a lot of sites (esp. any site with cache-able content thats normally
quite busy or is on a slower uplink than your downlink, etc.).

-- 
								Greg A. Woods

+1 416 218-0098;  <gwoods at acm.org>;  <g.a.woods at ieee.org>;  <woods at robohack.ca>
Planix, Inc. <woods at planix.com>; VE3TCP; Secrets of the Weird <woods at weird.com>