Re: Regarding polygraph useage

From: Alex Rousskov (rousskov@measurement-factory.com)
Date: Thu Feb 22 2001 - 06:33:30 MST


Venkatesh,

        PolyMix-1 is an ancient workload. Why on Earth did you decide
to use it? Please use PolyMix-3 instead:
        http://cacheoff.ircache.net/N03/
        http://www.measurement-factory.com/docs/PolyMix-3/

Before you run a PolyMix-3 test, you may want to find an approximate
peak request rate your box can support. A "boiling frog" workload with
steadily increasing request rate over a few hours may be helpful. If
you cannot build such a workload and cannot find its description in
the e-list archives, please contact me off-list.

I do not know how well your Squid box should perform under PolyMix-3.
You can get some very rough approximation by comparing your box/Squid
configuration with the ones tested at the last cache-off.

Good luck,

Alex.

P.S. IIRC, PolyMix-1 has only one parameter -- peak request rate.

On Thu, 22 Feb 2001, gv_squid wrote:

> Hi Alex,
>
> We are in the processes of testing squid(2.3STABLE4) using
> your benchmarking tool polygraph-2.5.4.
>
> This is what we did in our own bake-off,
>
> We fixed the maximum request per second as 25.As given in
> polygraph document we spilitted the max request in to 10
> (3,25,5,22,8,20,10,17,13,15).Each of these 10 test have been
> carried out for a hour.
>
> As a result squid was not able to handle more than 17 request.
>
> For the above test we followed Polygraph documentation.
>
> We changed only the req/sec and time_duration directives in
> polymix-1.pg(For all the ten tests).
>
> Is there any other directives to be changed in polymix-1.pg? What
> is the purpose of polymix-2.pg and polymix-3.pg?
>
> Running each test for a hour is enough or we have to increase the
> time-duration?
>
> Thanks for any advice.
>
> Venkatesh.
>
>
>
>
>



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