RE: multiple squid with load balance

On Fri, 9 Jul 1999, Jan Torreele wrote:

> can you provide some more information about these bizarre behaviours/poor
> performances you're seeing with L4 switches (webpage maybe?)

Not yet. The observations are a part of our Data Communications testing:
        http://polygraph.ircache.net/Events/datacomm.html
The results are scheduled for publication in September.

If there is enough demand and support, we may conduct dedicated L4
tests. I do not expect that happening within the next month though. That
is why I suggested custom, on-site tests. Despite simplistic workloads
(as far as L4 switching is concerned), Polygraph is quite effective for
such tests, and workloads are getting better.

> Also, what kind of loads are you talking about? 10Mbit/s? 100Mbit/s?

Mostly around 100Mbps and higher, but some bugs are present at lower
rates as well.

I regret that we cannot provide any specific data at this point. The
purpose of my message was to warn people not to perceive (by default) a
L4 switch as something that works as reliably and transparent as a L2
switch.

Break it before you buy it.

Alex.
Received on Fri Jul 09 1999 - 11:16:21 MDT