[asterisk-biz] Maximum Load
Matthew Rubenstein
email at mattruby.com
Wed Jul 18 10:03:37 CDT 2007
No one can answer that frequent question with a definite answer. There
are more variables than whether Asterisk handles the media, though that
is important. Transcoding per call, legs per call (conference), whether
it's recorded to disk, ramdisk/network strategies, all multiply or
divide the load per "call". Plus, server HW varies quite a lot more than
some virtual "box".
Digium should perform a large test project on various kinds of popular
hardware, and some optimized (per config) hardware, and publish the
benchmarks. It would be good to know how many calls, with no
transcoding, recording or any other bells & whistles other than just
completing two legs, incoming and outgoing. On a single minimum HW
config of, say, a generic P3/2.3GHz/1GB/100Mbps box, a
P4/3.2GHz/1GB/100Mbs box, a 4x-Xeon/etc, and maybe a
P2/500MHz/512MB/100Mbps notebook or something. Then give some estimates
of how many MHz each extra operation (like each leg transcoding, or
recording) consumes on each base HW config. And a formula for combining
the features in a given Asterisk deployment into an estimate that can be
used for capacity planning, including the overhead for Asterisk to load
balance in a cluster.
A similar benchmarks table for OpenSER, both standalone and fronting
for Asterisk, on those *identical HW configs* would be good. Or just a
*quantitative* comparison of Asterisk capacity vs Asterisk + OpenSER (on
two boxes). Though OpenSER doesn't have the corporate ($) organization,
with a marketing team ($), that Asterisk has to complete that, it's a
simpler benchmark process because OpenSER has so few combinations of
configs. The OpenSER community can probably do it, after Digium
publishes an Asterisk table for it to augment.
These are the kinds of basic benchmarks that any serious server product
needs to publish. Without it, every developer and deployer is operating
in the dark, by the seat of our pants. Each delivery requires that
testing, so time and money is spent on a repetitive operation that could
be done once, *by Digium, the vendor*. When competing with other
servers, the more professional organizations will offer the benchmarks,
while Asterisk has a harder time competing without them, or with a cost
(including delay) to produce them, while looking less professional than
"official" benchmarks.
It seems to me that this kind of benchmarking project would be a great
way for Digium to train people about Asterisk. So Digium could do it at
a training seminar or two. Especially now that Digium owns AstriCon, and
Adtran owns both of them, I don't know why Digium would wait any longer
to produce these valuable marketing/engineering data.
How about it, Digium? Help us to sell your products into more serious
corporate environments.
On Tue, 2007-07-17 at 23:19 -0600, Seysan wrote:
> Hi Everyone,
>
> I want to know how many concurrent calls can asterisk handle on a
> Single Server BOX.
> How many in Proxy State mode ? (that media passes through Asterisk)
> and How many in Signaling mode ?
>
> I think with a descent system, for an Office it can easily handle at
> least 50 phone calls at the same time, but I want to know how many in
> situation like Calling Cards or as a Softswitch like operation for
> wholesale traffic.
>
> Is OpenSER or SER is better for wholesale and calling card ?
>
> Thanks to all.
>
>
> _______________________________________________
> --Bandwidth and Colocation Provided by http://www.api-digital.com--
>
> asterisk-biz mailing list
> To UNSUBSCRIBE or update options visit:
> http://lists.digium.com/mailman/listinfo/asterisk-biz
--
(C) Matthew Rubenstein
More information about the asterisk-biz
mailing list