[Asterisk-Dev] Is there a need for a DS3 channel/driver?

[John Todd]

Following up on this rather long thread, I'll reply to the first 
message instead of the thread fragments which have gone in too many 
directions.  As a background note, there was extensive commentary on 
this on the -users list around a year ago 
(http://lists.digium.com/pipermail/asterisk-users/2003-December/029309.html)

Comments on this thread:

1)
   There are cards that handle channelized DS3's, to whoever it was 
that commented about unchannelized DS3 hardware. 
http://www.sbei.net/Products/WAN/wanPMC-C1T3.htm  is a good example 
(and yes, they can include a PCI adapter.)  This device includes 
Linux drivers for RH.  I don't know anything about implementation, 
driver compatibility, etc. - I just know it's a channelized-to-DS0 
card, and with the PCI carrier it's about $3500, quantity 1.  Perhaps 
there are cheaper cards?


2)
   No, none of these cards to my knowledge include DSP resources for 
codec transcoding, which is the "heavy lifting" numerical calculation 
problem.  However, many times codec transcoding is neither required 
nor desired, so this is not a major hurdle, and there are other 
solutions for offloading of computation.  In places where transcoding 
is required, IAX2 or SIP or TDMoE can hand off to number-crunching 
machines which have the duty of interacting with distant 
codec-constrained endpoints.  In fact, you almost never want your 
gateway device to be the system that end-users interact with directly 
from the VoIP side, anyway, but that's a different discussion.

3)
   There is a business case here, but it's a niche, certainly.  The 
argument for TNTs is fine, but they're still ~$10k, even used (though 
I'm sure that you can find them here and there for cheaper - this is 
not a price discussion, and comments of "I've found it for $3.22 on 
eBay!" will be ignored.)  A solution with a DS3 card that fits into 
1u is extremely interesting from a space saving perspective; larger 
cabinet footprints have a price associated with them, and while there 
may be people that get cheap DS3 loops, I know that most focus right 
now is on datacenter cross-connects, where the price of the cabinet 
space is usually about 1:3 to the price of the cross-connect, so 
space savings is Good.  Having a system that is extremely flexible 
within Asterisk's current protocol selection is also desirable, 
reducing the cost of administration.  (H.323? SIP? SCCP?  IAX2?!?) 
The true scalability of IAX2 might shine very brightly in trunked 
situations with this volume of channels, saving significant bandwidth 
costs between endpoint locations.

   The comments about leaving this to the telcos: uh, this is FOR the 
telcos.   This is not for most ITSPs, who would be hard-pressed to 
come up with 672 simultaneous channels anyway.  (To those of you that 
do have 672 simultaneous channels: Congratulations!  But you're in 
the minority, so don't feel like you need to object in a response to 
this post.)   Indeed, many larger carriers are offering SIP handoff 
these days, and the number grows every quarter - it is an admirable 
trend.  However, simply because there is a trend towards moving away 
from DS3 interfaces does not mean that there is no market for an less 
expensive DS3 gateway.  We're a long way from getting rid of TDM, 
folks, despite our evangelism of VoIP - you can ignore that fact, or 
you can make money on that fact while also embracing the new 
technology of packet-based call delivery for even cheaper call 
delivery.  Choose your camp.


4)
   Regulatory issues: most card vendors have regulatory approval, at 
least in North America.  Other nations have the typical red-tape 
issues, so outside the US this may be of limited use, especially in 
nations in Asia where both software and hardware are more stringently 
tested in combination.   Still requires examination.


5)
   SS7: Yes, SS7 is still a problem with Asterisk.  Moving to DS3 
interfaces just makes it more painfully obvious.  NFAS PRI is where I 
would see this working first, and then ISUP eventually.   I think 
Race's proposed suggestion would be for fairly dumb TDM-to-VoIP 
conversions, and not a Class 4 or Class 5 switch extension.  It's 
just to get bigger density of ports, and we shouldn't look at it as a 
simultaneous expansion into SS7, or it will never happen.


6)
   I've said before, and I'll say again: most of the biggest users of 
Asterisk never appear on this forum.  They never say anything.  You 
don't know they're using Asterisk.  They're reading this thread 
closely, but due to various NDAs, competitive market nonsense, and 
outright FUD, they won't comment on it.  My suspicion is that 
Asterisk is already running DS3 cards somewhere inside several large 
telecommunications firms, but that information will never leak back 
to us in any public manner.  This is neither good nor bad, it's just 
a part of Open Source that we have to deal with.  If there is 
interest from one of these large firms on the DS3 project, maybe 
they'll step up with some funding for Race, but maybe not - if the 
OSS foundations of the project means that they can't keep a 
competitive edge (or if they're not able to explain it to their boss) 
then they won't back the project.

   IBM: I'm waiting.  Step up to the plate, guys - this, along with 
many other Asterisk improvements, is where you could make a 
difference in telco-grade Linux.  Same with OSDL.org - there seems to 
be a no focus on applications, despite that being where the greatest 
changes could happen for market uptake.

JT