Re: Squid & Netscape Proxy

From: Ong Beng Hui <ongbh@dont-contact.us>
Date: Sat, 25 Jan 1997 09:40:53 +0800 (SST)

> > This is interesting. The UK National Cache currently runs Netscape
> > Proxy (I'm in the process of planning a possible change to Squid). We
> > have a monitoring script that checks each machine every 5 minutes to
> > see that the cache is OK. We frequently get errors stating that
> > requests timed out at 600 seconds (10 minutes). It looks like there's
> > some sort of problem in Netscape proxy, and we aren't the only people
> > to experience it.
>
> Netscape Proxy doesn't support ICP (nor will will it - note this is a challenge
> to their developers!).

I guess another problem with Netscape proxy will be it's
multiple process forking design. I am not sure if Netscape
Proxy v2.5 does threading within a single process, but for
v2, it doesn't appear so.

So, it is fairly easy to calculate how much load can your
server handle by dividing your memory size with the process
size. I don't think Netscape store the complete index in
memory, although there seems to be two GC process hanging
around.

It is interesting to see how UK National Cache can handle
it's load with Netscape Proxy on SGI.

I like Squid (Harvest) single process non blocking design,
but it will be nice to does multithreading within it's
design. It is also difficult to scale a Squid Cache in a
multiple CPU server because it is a single process design.

I am currently running Harvest V2 on a 4 CPUs DEC 8200 5/300.
The only reasonable thing I can do will be to divide the
total disk storage into 3, and run 3 Harvest process. The
OS will typical run each of those process on different
CPU. If I add additional CPUs, I will have to run more process
and further divide the disk. And that's not very scalable.

I am sure there are a number of Squid users sharing the
similar problem on SUN server that can probably support a
dozen CPUs.

*8)
Ong Beng Hui
ongbh@singnet.com.sg
...yet another day in an ISP business
...and they lived happily ever after
Received on Fri Jan 24 1997 - 17:50:05 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:34:10 MST