Re: proxy cache cooperation

From: Alex Rousskov (rousskov@measurement-factory.com)
Date: Thu Feb 15 2001 - 19:46:41 MST


On Thu, 15 Feb 2001, Ping Du wrote:

> Do you think it is possible to test performance of proxy
> cooperation by running Polygraph? For example, I have two proxy
> caches and two poly server/client pairs, one client is pointed to
> one proxy, the other client is point to the other proxy. Moreover,
> the two proxy caches know existence of another proxy server by
> some ICP configuration. Can I use normal Polygraph workload this
> way?

Yes, recent standard workloads such as PolyMix-2,3 can be used in your
scenario.

> Is there any potential problem here?

Make sure that all polyclt and polysrv processes are using the same
PGL configuration. Also check that Robots' public_interest is
positive. It is positive in standard workloads. This will ensure that
Robots from different machines will share the specified portion of the
URL space, making peer hits possible.

You may also need to play with the working set size and popularity
model to get the desired effect (i.e., more/less peer hits and
misses).

I only used load balancing approach to aggregate proxies, so I may be
missing some important points. Depending on your testing goals, you
may have to adjust standard workloads.

Alex.



This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:17 MDT