[Asterisk-Dev] rewriting asterisk as a state machine? - Not

[Michael Giagnocavo]

Not sure, but it seemed like people doing huge numbers of calls on a single
machine are doing it pure IP with codec passthrough -- not to zap.

At what point did your system still have a bit of CPU free and cause no HDLC
errors at all?

-Michael

-----Original Message-----
From: asterisk-dev-bounces at lists.digium.com
[mailto:asterisk-dev-bounces at lists.digium.com] On Behalf Of Chris A. Icide
Sent: Monday, March 07, 2005 1:01 PM
To: Asterisk Developers Mailing List
Subject: Re: [Asterisk-Dev] rewriting asterisk as a state machine? - Not


-----BEGIN PGP SIGNED MESSAGE-----

Was there ever a response to this as to what system design (I assume
the 
poster was referring to asterisk since this is an asterisk mailing
list) 
was capable of handling 5000 concurrent calls?  And what did the
caller 
mean by 5000 concurrent calls.  Are these SIP to SIP, SIP to IAX, SIP
to ZAP?


I recently was able to run some tests on call density myself.

I was using a Dual Processor Xeon 3.06GHz system with 4 4-port T1
cards 
installed (for a total of 16 t1 ports)

I had two of these machines with cross-over cables connecting all 16
ports 
with each other.

I then implemented a simple test in which I connected a call from one
machine to the other across the T1 interface terminating to a
Milliwatt on 
each end.

Running a Linux 2.6.10 kernel, I was able to fill all 368 channels
with 
echo canceling disabled.

According to top, with all channels filled, the CPU was running at
99.9% 
for the asterisk process.


I also tried the test with echo cancelling enabled, but upon hitting
the 
10th T1 of traffic, I got a flood of HDLC errors (cpu usage hit
99.9%) and 
calls were dropped.

In the no echo cancel test, with all channels filled, I was getting 
sporadic HDLC errors.

5000 concurrent calls seems interesting.  I suspect the method of
achieving 
this is through an army of systems sharing load?

- -Chris

On 05:51 AM 3/4/2005, Brian West wrote:
 >
 >I would like to know too... since it was only possible to get about
185
 >or so threads till very very recently.
 >
 >/b
 >
 >On Mar 4, 2005, at 9:14 AM, Andrew Kohlsmith wrote:
 >
 >> On March 4, 2005 09:52 am, Paul Mahler wrote:
 >>> We are getting over 5000 simultaneous calls with our PBX
hardware
 >>> with less
 >>> than 50% CPU utilization, so I'm not sure this is much of a
problem.
 >>> ;-)
 >>
 >> What hardware and what types (channels) of calls?  I'm just
curious,
 >> especially as to what you do for redundancy and failover.
 >>
 >> -A.
 >> _______________________________________________
 >> Asterisk-Dev mailing list
 >> Asterisk-Dev at lists.digium.com
 >> http://lists.digium.com/mailman/listinfo/asterisk-dev
 >> To UNSUBSCRIBE or update options visit:
 >>    http://lists.digium.com/mailman/listinfo/asterisk-dev
 >
 >_______________________________________________
 >Asterisk-Dev mailing list
 >Asterisk-Dev at lists.digium.com
 >http://lists.digium.com/mailman/listinfo/asterisk-dev
 >To UNSUBSCRIBE or update options visit:
 >   http://lists.digium.com/mailman/listinfo/asterisk-dev
 >
 >
 > 

-----BEGIN PGP SIGNATURE-----
Version: PGP 8.1

iQCVAwUBQiylAO0LTNca2q41AQGS0QQA2PFkDCVGDAQjVxXTKCKuiFTdSrptsCw8
QYu0mdMpkgz0U8Dn6PfaAA+S6PKNI7oy9dYVXuAdXogQuYljY/0GGI+OeNSu+Uvm
zpWAyL1ObUtC7rQBh4cKvWUOCGv2rqDv96SfuxqailqgHaKxvPeLte71ziwZUdy2
ZjCvSddEKzw=
=xKmD
-----END PGP SIGNATURE-----

_______________________________________________
Asterisk-Dev mailing list
Asterisk-Dev at lists.digium.com
http://lists.digium.com/mailman/listinfo/asterisk-dev
To UNSUBSCRIBE or update options visit:
   http://lists.digium.com/mailman/listinfo/asterisk-dev