On Wed, 24 Nov 1999, Alan Smithee wrote:
> > Note that WebAxe-1 requires phase-dependent
> > hit ratio, a feature not yet supported by Polygraph.
>
> Any comments on why this is an issue for reverse proxies?
Here is why we propose to vary the recurrence ratio during reverse proxy
tests.
The URL space of most Web sites does not change much so the recurrence
ratio should be close to 100% during the measurement phase of the test
(we propose 95%). We also need to bring the proxy into a steady state by
populating its cache with objects. The latter can be done with 95%
recurrence ratio, but would require significantly longer experiments
because of the very low fill rate (less than 5% of the request rate).
To shorten the duration of an experiment, we populate the cache with 5%
recurrence ratio and then switch to 95% during the measurement phase.
Alex.
P.S. The offered hit ratio is the recurrence ratio adjusted for
non-cachable objects.
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:09 MDT