[Asterisk-Users] PRI issues with TE410P

[Storer, Darren]

Hi Azher,

> They advertised on TeleVision and then we had a rush of
> calls landing on the system (about 30 calls in 1-2 secs and 60 in 3-4
> seconds time).

this sounds like a high number of _simultaneous_ call attempts for a PRI
connected system to support. Your comments about a gradual load increase
presenting no problem would seem to back this fact up. On most Public
Telephone Switches (in the UK) it is advised that no more than 5 call
attempts per second should be attempted via PRI if you seek stability.
(Above 5 calls attempts per second "Network Congestion" messages can be
sometimes be observed on a protocol analyser. These congestion messages can
prove fatal for certain applications such as high quality video conferences
via hardware that aggregates multiple B channels.) It should be remembered
that PRI connections are based upon subscriber signalling specifications and
that "spiky" mass call events are traditionally hosted on platforms that are
connected via SS7 signalling.

I know of similar stability problems with Dialogic and Acculab ISDN
connected systems where mass call events, in excess of 5 call attempts per
second, have proved too much of a spike for the Public Switch. In certain
switch architectures concentrator equipment can also be adversely affected
by this kind of deployment which can cause disruption for many customers
(not just the one running the PRI IVR platform).

On slower CPU systems (< 800 MHz) I have noticed occasional Euro ISDN PRI
signalling problems if a high percentage of the 30 channels on an E100P are
busy with simultaneous outbound/inbound calls and glare occurs. On the
protocol analyser trace it was evident that some messages via the D channel
from the Public Switch went unacknowledged by the Asterisk system. If you
encounter this problem a quick work around is to split the channels of your
single PRI into two trunks to stop the contention between inbound and
outbound telephony traffic. (E.g. g1 for incoming and g2 for outgoing calls)

I have yet to flex a TE410P with significant traffic levels as all my busy
systems are using E100P cards but I have just finished commissioning a
TE410P system (with help from Rob Boardman - thanks) in Munich that will be
stress tested over the next two weeks; I hope to have news on stability when
the tests are completed.

HTH

Darren
--
Comgate (UK)
Telco>Internet<Broadcast

To:  <asterisk-users at lists.digium.com>
Subject:  RE: [Asterisk-Users] PRI issues with TE410P
From:  "Azher Amin" <azher at consulttech.com.pk>
Date:  Thu, 25 Mar 2004 09:09:18 +0500

Hi Tan,

I have already turned off the HT from the BIOS (while it was enabled,
asterisk was not even stable with 2 pri's).

Further I am using the same kernel (2.4.25) with SMP enabled, that u
mentioned in your email.

Basically I think Scott is right in his emails, that libpri's and zaptel
driver for t4xxp are not really optimized and they really need tunings
in case when there is certain increase in number of calls ....  Also I
cannot afford to get a system higher than the current specs.

Yesterday I had the same situation, I gave the IVR solution using te410p
to a company. They advertised on TeleVision and then we had a rush of
calls landing on the system (about 30 calls in 1-2 secs and 60 in 3-4
seconds time). What I have observed is:

1. Asterisk is stable upto 2 PRI's ... no call drops no errors.
2. As soon as I plug in the 3rd pri (3rd pri is confirmed stable), it
starts dropping the calls current landing as well as the previous stable
landed calls. All calls get dropped and PRIs' start syncing / unsyncing.
3. On the other hand I have also a dialogic server with 4 pri's in it,
and it was a rock solid ... going to full 120 calls ... btw: server is
just P3 700Mhz, and I have not noticed a single dropped call.
4. My configurations were checked by Mike (@ digium) ... and according
to him all things are fine, zaptel,zapata,extensions.conf ...

Thus when there is a smooth and gradual load of upto 3 pri's system will
remain stable, and when there is an instant call load it becomes shaky.

Also I will request those who have successfully done similar setup, to
post/advise their linux flavor (and updates), asterisk configurations,
kernel configuration, other system tweakings.

I want Mark from Digium to take special interest in solving and
recommendation on these issues, coz I have confirmed orders of 3 te410p
boards and one e100p (for which I mailed to Greg as well), but now I am
feeling reluctant to face the client.

Regards
Azher Amin
-------------------------------
http://www.consulttech.com.pk


-----Original Message-----
From: asterisk-users-admin at lists.digium.com
[mailto:asterisk-users-admin at lists.digium.com] On Behalf Of
tan at yointernet.com
Sent: Wednesday, March 24, 2004 5:03 PM
To: asterisk-users at lists.digium.com
Subject: RE: [Asterisk-Users] PRI issues with TE410P

Before going to gentoo or another version of linux, try the following:

1) turn hyperthreading off in the BIOS. It's probably called something
like "virtual cpus".

2) Use a vanilla kernel from kernel.org e.g. 2.4.25.

We used to have times when the system would just hang. Using a
non-redhat kernel seems to have resolved that. We also use Gentoo, but I
warn you that it is not the easiest flavour to set up!

Hope that helps.

Tan
Telappliant.com
Voiptalk.org