Re: Squid 1.2 Release Date & Effectiveness of Proxies

On Mon, 20 Jul 1998, Dancer wrote:

> Tim Brody wrote:
> >
> > ( ) -- ( )
> > ( ) / o \ ( )
> > ( )__ \ / ( )__
> > (____) --- (____)
> >
> > More seriously, I need some ammunition. I want to install a proxy at my
> > local ISP (I'm sure the enthusiasm should be the other way round), but they
> > think it's a waste of time..."it's just another machine for the request to
> > go through".
> >
> > Can anybody give me some _general_ information on the benefits of Proxies,
> > like do Proxies (squid1.2) slow down requests, how effective is a proxy on a
> > low bandwidth connection, i.e. ISDN? Do proxies pose any security risk,
> > assuming that ACLs block external access to the proxy. Most information on
> > proxies seem to cover T1/T3 type load ... I'm not really in that ball-park!
> >
>
> Yes, it works _well_ over ISDN connections. Impressively so, except (of
> course) that ICP tends to get lost when things get busy...so you can't
> really use ICP between multiple parents. Squid loses sight of them when
> things busy up.
>
> I've got a bunch of clients now all using squid over ISDN. Think: '40%
> more delivered data to customers for the same cost as before.' The
> percentage varies on setup, but any piece of software that will do that
> for you is a win, IMO.
>
> Also, two shorter data-paths (or three, even) between two widely
> separated points are less likely to break from congestion than a single
> link between the endpoints, and much liklier to show higher data rates:
> The effects of RTT on end-to-end bandwidth. Even if there's a little
> postage-and-handling at each proxy in the chain, or slightly increased
> times in starting the fetch of an object, the delivery time tends to be
> much improved and the delivery itself tends to be more reliable.
>
> D
>
> --
> -----BEGIN GEEK CODE BLOCK-----
> Version: 3.1
> GAT d- s++: a C++++$ UL++++B+++S+++C++H++U++V+++$ P+++$ L+++ E-
> W+++(--)$ N++ w++$>--- t+ 5++ X+() R+ tv b++++ DI+++ e- h-@
> ------END GEEK CODE BLOCK------
>
>

I have proxy users that have reported quicker response times even
on cache misses and here is my theory as to why. The user is dialed
up via a modem which is slow. Me, the ISP has a link that is fast
by comparison. Connections to any given site on the internet gives
thruput that is frustratingly sporadic. You get a couple of seconds
of flow, then stall for a couple of seconds. The cacheing server
connected via a faster link, is going to get more data in
that couple of seconds of flow that the end user over his little
modem line. As a result, a sporadic trickle for the cacheing
server may be able to keep the end users modem line saturated.
Thus the end user is happier.

It does seem to cause bandwidth usage to go until there is enough
stuff in the cache to give a significant hit rate.

-

Barrett Richardson rabtter@orion.aye.net
Received on Sun Jul 19 1998 - 18:52:10 MDT