Re: Stressing in cluster??

From: Alex Rousskov (rousskov@measurement-factory.com)
Date: Tue Feb 06 2001 - 08:17:26 MST


On Tue, 6 Feb 2001, Mulyadi wrote:

> Is it possible to spawn several instance of server and client with
> different scenario and distribute them in cluster.

Sure. During cache-offs, for example, each client-server pair is
configured to generate at most 400 req/sec; we add more pairs as
needed to get the desired peak request rate. See cache-off reports
for details.

Standard workloads are symmetrical and "uniform". That is, each
Polygraph client (server) runs the same number of equally configured
robot (server) agents. One can design a workload that is not
symmetrical and not uniform. Most configuration options are per-agent.

> I want to do it because i think it will speed up the test, any
> suggestion or critics??

What is "speed of the test"? The total peak throughput should be
determined by the device under test. Polygraph can accommodate
virtually any request rate by adding more client and/or server PCs. As
a rule of thumb, make each Polygraph PC comfortably loaded and add
more PCs as needed to support that level of "comfort" while getting
the required total request rate.

Alex.



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