Re: Usage of ProxyCacheSize in cooperative proxy caching test

From: Alex Rousskov <rousskov_at_measurement-factory.com>
Date: Thu, 31 Oct 2002 10:12:49 -0700 (MST)

On Thu, 31 Oct 2002, Ko Tensei_at_KIT wrote:

> Maybe a fundermental problem only this time. It's about matters that
> require attention when setting a cooperative proxy caching benchmark
> enviroment using Web Polygraph.
>
> It's clear that ProxyCacheSize represents "total cache size" in
> terms of hd and memory when testing a single cache server, however,
> when testing multiple cache server, should this parameter be set to
> the total ProxyCacheSize of all cache servers?

ProxyCacheSize represents "total cache size". If your cache consists
of several components (distributed or not), all of them have to be
accounted for.

> ProxyCacheSize is used only in fill phase. As in polymix-4-guts.pg we
> can see, phFill.goal.fill_size = 2 * ProxyCacheSize / clientHostCount,
> (where clientHostCount=TheBench.client_side.hosts ).
>
> In my case I have 3 cache servers with a total cache size of say 803MB,
> and 1 polysrv process, 2 polyclt processes(client_side.hosts=['...','...']).
> So when I set ProxyCacheSize to 803MB,
> phFill.goal.fill_size should be 2 * 803 / 2 = 803MB.
> However when I run the test and check srv.log and clt.10.101.0.log,
> they both say that phFill.goal.fill_size equals 1.57GB (2*803MB)
> -- clientHostCount is ignored?
> Furthermore, polysrv doesn't stop when it reaches 100% of the goal.size,
> but only stops at 200% of the goal instead.
> That means the actual goal.size will be 4 times of ProxyCacheSize?
> (WSS is limited to be smaller than goal.size in my case)

What you describe should not be happening. I would need to see your
workload files and console output to understand whether it is a bug or
misconfiguration. Please e-mail me off-list if you are still having
these problems.

> Before checking the source, I need a confirmation on the usage of
> ProxyCacheSize in cooperative proxy caching benchmark, will u please
> give me any suggestion?

PolyMix is for testing a single cache "system". From PolyMix point of
view, you have a single cache (with 3 distributed and cooperating
components). To comply with PolyMix rules, you have to flush all your
caches before the test and fill all of them twice (pipe enough fill
responses through the cache system) before proceeding with the
measurement phase.

Alex.

-- 
                            | HTTP performance - Web Polygraph benchmark
www.measurement-factory.com | HTTP compliance+ - Co-Advisor test suite
                            | all of the above - PolyBox appliance
Received on Thu Oct 31 2002 - 10:54:28 MST

This archive was generated by hypermail 2.2.0 : Fri Nov 21 2008 - 12:00:08 MST