On Thu, 12 Apr 2001, venkateswaran wrote:
> I am using polymix-1.pg. Now i plan to use polymix-3.pg.
> I am using two machine each for client and server, in local
> I have just edited polimix-3.pg, polymix-3-guts.pg, benches.pg and
If you want to run a standard PolyMix-3 test, do not edit
polymix-3-guts.pg, benches.pg, or any files in the include/ directory.
All PolyMix-3 parameters should be set in polimix-3.pg.
> peak_req_rate = 10/sec;
> TheBench.peak_req_rate = 10/sec;
> size ProxyCacheSize = 1010MB;
> rate FillRate = 75%*TheBench.peak_req_rate;
> addr rbt_ips = ['172.16.1.114'];
> addr srv_ips = ['172.16.1.115:2020'];
> recurrence = 5%/HitIfRepeat;
> time platDur = 1hour;
> Is this enough?
This is way more than enough!
A PolyMix-3 robot designed to emit 0.4 req/sec and has a limit of 4
open connections. You manually specified one robot (at address
172.16.1.114) so your total peak request rate will probably not be
PolyMix-3-guts.pg file is designed to compute the number of robots and
the robot addresses based on peak_req_rate. If you manually set the
addresses, you are responsible for calculating them based on your
needs (and for all the bugs your calculation may have).
PolyMix-3 rules require at least 500 simulated servers. You manually
specified just one (at address 172.16.1.115:2020). PolyMix-3-guts.pg
file is designed to compute server addresses based on peak_req_rate.
You should not modify recurrence unless you want to have offered hit
ratio different from PolyMix-3 rules.
All of the PolyMix-3 parameters could and should be changed in
polimix-3.pg file only to avoid the bugs outlined above. You cannot
change phase durations in polimix-3.pg, so you will have to do that in
Polmix-3-guts.pg (but phase duration is not a PolyMix-3 parameter -- a
shorter test will not produce valid PolyMix-3 results).
> But i get difference in cache filling.
> For 1 hour duration for 1GB cache filling
> polygraph-1.pg has filled 856MB but
> polygraph-3.pg has filled 145MB
This is probably because you [unintentionally] configured Polygraph to
have just one robot. Note that PolyMix-1 required a separate
filling-the-cache test. PolyMix-3 has integrated fill phase. Also,
PolyMix-1 robots have different dependencies on workload parameters so
it is possible that given two misconfigured tests (polymix-1 and
polymix-3) one will "outperform" the other.
Standard workloads will perform correctly. If you modify the workloads
to an extent that they are no longer "standard", then the side effects
will depend on your modifications. This situation is similar to
driving a factory car versus driving a car with the engine you
designed yourself: both are possible, but the former is much more
likely to produce predictable results.
> And after 1 hour both client and server does not stop.
PolyMix-3 fill phase stops when the cache is filled twice, not after
one hour. The plateau phase duration applies to top1 and top2 phases
> please tell the procudure for quickstart for ploymix-3.pg.
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:18 MDT