On Tue, 20 Jun 2000, Ittai Gilat wrote:
> I have some questions regarding polygraph 2.5.0 and polymix-3:
> - Do I have to upgrade to FreeBSD 3.4?
No, but we will not use FreeBSD 3.3 for the tests. You must upgrade if
you want to to be safe from any surprises the difference in OS versions
might bring.
> - If I don't have to upgrade, are there any changes that I have to do to my
> current FreeBSD 3.3 machines that ran polymix-2 on polygraph 2.2.9?
I assume you have already applied PolyMix-2 changes. There is a kernel
bug in FreeBSD 3.3 that you probably want to fix. See
http://polygraph.ircache.net/cgi-bin/mail2html.cgi?fnames=20000601/712
> - I noticed that polymix-3 has the fill-cache phase included. Am I correct
> to assume that this means that the working set is full by the time top1 is
> entered?
Not exactly. WSS is specified using time (e.g., 4 hours). The duration
of the filling phases is unknown a priory as it depends on the proxy
cache size and fill rate. Some proxies take more than 4 hours to fill,
some take less than an hour!
> If so, what is the need for top2 (which is another 5.5 hours)?
Top2 is actually 4 hours (5 with the ramp phases). We need Top2 to check
whether the proxy was in a steady state during Top1. If you look at the
last bake-off results, you will notice that many products did not reach
their steady state in Top1. Full cache is not the only factor that
contributes to the steady state conditions.
For your private tests, you can, of course, shorten the duration of the
experiment if you are positive that the proxy under the test reaches
steady state during Top1. Official tests would have to run longer.
However, if we get sufficient number of reports saying that steady state
conditions are reached in Top1, we will certainly re-evaluate phase
durations. Clearly, there is no reason to run a longer test if the
results remain the same. So far that was not the case.
Thanks,
Alex.
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:13 MDT