RE: [xmlblaster] Durable QoS performance

[Doug.Palmer at csiro.au]

> Great, still the performance is not overwhelming, ist your CPU load 
> still as low
> as you mentioned in one of your previous mail?

For non-durable at maximum throughput using oids, The CPU ranges from 60-80%
for both CPUs, the Disk 0% and the number of IP datagrams is 6000-9000 per
second.
For durable at maximum throughput using oids, The CPU ranges from 60-90% for
both CPUs, the Disk 0-20%[*] and the number of IP datagrams is 2500-3000 per
second.

This is (kind of) respectible, given the number of IIOP, RMI and whathaveyou
network messages flying about.

[*] This shows periods of 0% interspersed with spikes. I would guess that
this is where xindice flushes its buffers to disk. But this would mean that
using xindice as a persistent store may result in lost messages. I'll
investigate further.

> Could you explain what the columns mean exactly?

Oops.

Size  Approximate message body size in bytes   
MPS   Attempted total send speed in messages per second  
Sndrs Number of senders   
Rcvrs Number of receivers
Grps  Number of "groups" (subject names)
Time  Time spent running the test in ms
Sent  Number of messages sent
Rcvd  Number of messages received
STpt  Send throughput in messages/s
RTpt  Receive throughput in messages/s

If there is more than one group, then senders and receivers are distributed,
more or less evenly, between them.