I had the same problem. Apparently, there was a problem of time
co-ordination among the machines. I installed an NTP software to have
all the machines keeping track of the same clock, and client and server
moved from one phase to the next almost at the same instant.
Using only one client and one server takes quite effort from those
machines, and if clocks are not very good, perhaps the OS has better
things to do than keeping the clock adjusted, like managing buffers and
connections (quite a lot).
You can get NTP at http://www.eecis.udel.edu/~ntp/
Good luck
"Jihong A. Kim" wrote:
>
> Hi,
>
> Q1)
> Is it okay to distribute IP address unevenly among clients or servers and
> run the polymix test? ( For instance, I would give one client machine 300
> IP addresses and the other 500 IP address, and run the test. )
>
> Q2)
> Do client machines and server machines have to move together at the
> same time from one phase to the next throughout, the Polymix-3 test?
>
> I ran the Polymix-3 test with 450 request/sec, with one pair of
> server/client machine. However, the client and the server were moving
> through a series of phases at a different rate (even when they were
> synchronized with local time server.). So when the client was finished,
> the server was in "top1" phase. The order of phases is as follows.
>
> ----------------
> warm -> ramp -> fill -> exit -> inc1 -> top1 -> dec1 -> idle -> inc2 ->
> top2 -> dec2 -> cool
> ----------------
>
> I'd appreciate any help.
>
> Thanks a lot in advance. =)
>
> - jihong
-- Luis ****
This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:19 MST