[openamq-dev] Tuning

[R Hayes]

The benchmark sends 100 messages as fast as it can, pauses for 1  
second, and repeats.  We run it for 100 of these cycles.  The results  
were consistent from run to run once we tamed the O/S (low latency  
kernel).

I'm on vacation now, but when I get back I will show you the data.

-r



On Oct 25, 2007, at 7:35 AM, Martin Sustrik wrote:

> Hi,
>
>> That was a 34% improvement on relatively exotic network hardware.
>> Presumably the percentage gain will be smaller for cases where the
>> CPU is a larger part of the overall latency.  Nonetheless, it can
>> have a big impact.
>
> I've did a single ping-pong test (bouncing-message back and forth) for
> malloc and tcmalloc (linux on dual-core box) - in both cases the mean
> latency and the median latency were almost the same.
>
> Can you elaborate a bit on what kind of testing scenario you've used  
> for
>  measuring the latency and - possibly - what kind of hardware have you
> used?
>
> Thanks.
> Martin
> _______________________________________________
> openamq-dev mailing list
> openamq-dev at lists.openamq.org
> http://lists.openamq.org/mailman/listinfo/openamq-dev