[asterisk-biz] Asterisk Performance Results

[Matthew Rubenstein]

	Ultimately a standard test suite like the one so helpfully published in
this thread would benchmark the baseline config, then the transcoding
config, as this one did. But for each of the combinations of the various
codecs. Then a benchmark adding each of various options, like
conferencing, recording, etc. That grid would be repeated on each of
directly comparable HW configs, like a single CPU with single core at
x-GHz, multiple of those CPUs, the same benchmarks repeated for
single/multiple multicore CPUs, each at increasing GHz.

	That 3D stack of benchmark data could be crunched to find the linear
(or other simple formula) ranges for scaling Asterisk capacity for
different featuresets, and where they are nonlinear or truly complex as
featuresets interact. A proper analysis could report basic units of
capacity that might be interchanged (if not too complexly interrelated)
like "adding recording to a call means dropping 10 2-party calls or 5
conference legs".

	That's the kind of benchmarking that I'd expect Digium to do. They
probably have already done at least a limited subset, but haven't
published any. In fact they decline to engage publicly in such a project
at all, saying it's too hard. But such benchmarking, to some degree of
completeness, isn't just a basic requirement for capacity planning in
deploying networks. It's also a guide to prioritizing work optimizing
the Asterisk code that's being benchmarked.

	At the very least I'd like to see some numbers that show exactly at
which scales adding scaling features like SER, RTPproxy, and even DUNDi
return extra capacity in exchange for their extra overhead, and the
actual Asterisk capacity ceilings at which they are necessary. Those
kinds of benchmarks are actually a perfect role for the Asterisk
community to collaborate on and share.

	Once this thread in -biz settles down, I'll see if I can get some
interest in the -users list. If someone else who can actually lead a
project by setting up a testing config wants to beat me to it, please be
my guest.


On Sat, 2007-11-17 at 16:33 +0000, Chris Bagnall wrote:
> > Asterisk was running on a server with two Xeon 5140, dual core, 2.33 GHz
> > CPUs and 4 GB of RAM.
> > We found that an Asterisk B2BUA on this hardware can manage 1500
> > simultaneous calls with no transcoding and 400 simultaneous calls with G.711
> > to G.729 transcoding.
> 
> Many thanks for the testing done so far. Hopefully it'll be of considerable use to the Asterisk community.
> 
> A possible future test:
> 
> What'd be very interesting to see is how Asterisk scales as hardware changes. For example, by what percentage do the call completion figures increase by switching to quad-core CPUs? Is it better to have slower quad-core chips, or faster dual-core chips?
> 
> Looking at the results, memory usage is fairly low, so one could reasonably deduce that 2GB RAM would be perfectly sufficient. Given that scenario, it'd be great to have some comparisons with different system specs to achieve the most "useful" system spec for a given budget.
> 
> Regards,
> 
> Chris
-- 

(C) Matthew Rubenstein