[rabbitmq-discuss] [erlang-questions] cheap low-power boards

Matthias Radestock matthias at lshift.net
Sat Jan 3 17:54:05 GMT 2009


Martin, Ben,

Ben Hood wrote:
>> I cannot proceed starting the performance testing client:
>>
>>        ./runperftest.sh 192.168.1.129
> 
> Sorry, I don't know how that script works (although it is in the java
> client package).

The script is tiny. All it does is invoke MulticastMain against a host 
specified as the first parameter, with a number of different parameters 
- which ones is what "running with ..." tells you. Each test is run for 
10 seconds and prints out the measured throughput.

> I don't know what " running with '  ' " means

See above. It just means that MulticastMain is run with the default options.

>> It would be great to get some reference results for comparison purposes
>> > as I am unable to interpret the attached results.
>> >
>> > Hope it helps the RabbitMQ project a bit.
> 
> Yes, you have started a useful discussion. I'm sure that some people
> in the Rabbit community can use your results to compare it with their
> current setups.

Comparing results of runperftest is only valid in a very limited sense.

There are inherent limitations in the test setup - the reported results 
do *not* represent the maximum throughput RabbiMQ is capable of 
delivering. You can't really read much at all into the figures, except 
that if you run the tests in the same setup against two different 
version of RabbitMQ and you see a significant (>10%) difference in the 
result then probably something has changed between the two versions :)

...which is exactly what runperftest was designed for: as a regression 
test that alerts us to changes in the performance profile.


There is a long outstanding task to create a test that really does 
measure the maximum throughput.


Regards,

Matthias.




More information about the rabbitmq-discuss mailing list