In webaxe workload the cache fill phase is processed at a request rate
of 75% of the top1/top2 phase.
Why is it so? In fact these 75% could cause a problem to the reverse
proxy machine: cache-hit being close to zero, the machine does much
more.
This could lead to a situation where the bottleneck of the test is the
fill phase and not the top phases.
A different policy was decided with polymix-3: the user could fix the
fill rate as he wishes. It seems much more logical since the purpose of
the test is to measure the top phases and not the fill phase.
I think there are good reasons to revise this issue and do the same as
in polymix-3.
Any thoughts?
Thanks,
Serge Ayoun
ISA Performance
Microsoft
This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:19 MST