Re: DHR (Document Hit Rate)

From: Alex Rousskov (rousskov@measurement-factory.com)
Date: Mon Apr 23 2001 - 10:45:43 MDT


On Thu, 12 Apr 2001, Sam Chen wrote:

> I am running Polygraph 2.5.4 to test our software. I am using
> the simple.pg. The original DHR (Document Hit Rate) from Polygraph
> Client output is about 40.0. After the polygraph runs for about 20
> minutes, one message shows up from Polygraph Server:
>
> client ran out of new public oids

Do you mean "from Polygraph client"?
        http://polygraph.ircache.net/doc/messages.html#_22
 
> then, the DHR starts to drift lower and lower. We let the system
> run for a long time, the DHR is just getting worse and worse to
> almost 0.

It is difficult to say without looking at the results, but I suspect
that there are other problems with the test such as excessively high
response time. Also, you may want to generate a report to check
whether _offered_ DHR is decreasing as well. If offered DHR does not
decrease, the proxy may be at fault.
 
> However, if we shut down Polygraph Server and restart it again,
> the DHR will jump back to about 40.0.
>
> Any suggestions on how to config simple.pg to make DHR stable at
> around 40.0 will be appreciated!

Please do not run production test with simple.pg. Use PolyMix-3 or
your own workloads based on PolyMix-3. Simple.pg is nearly the
simplest workload one can write with PGL; it is not meant to test
anything but basic network connectivity.

PolyMix has a lot of features keep request rate and offered hit ratio
constant (when needed) and that help to produce meaningful results. If
you find PolyMix-3 too complex (after following the HOWTO instructions
and spending at least some effort on trying to make it work), you can
try to simplify the workload or ask others to do it for you.

Alex.



This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:19 MDT