Slashdot Mirror
Is Linux Used in Production Telephony?
jamesva asks: "The telecommunications industry is rapidly converging on Windows NT/2000 for all telephony and voice-related needs. Most ACD systems, virtual operators, and voicemail are being ported to Windows if they're not already running on it.
In the past, telephony apps have existed most
notably on OS/2, SCO, and even DOS. However,
free Unix (or unix-like) platforms have absolutely no penetration in this area, with seemingly no chance on the horizon.
The Bayonne app server from the GNU folks seems to be the one exception, but even then there doesn't
seem to much built around it or anyone using it. It reached a 1.0 release in September and was met with no fanfare. Even the
LinuxTelephony doesn't seem to have much news. Can someone prove me wrong? Why is this the case? I'm interested in finding out if anyone is using Linux (or any free OS) in a production environment for something like voicemail or ACD. These types of systems require high availability and reliability and Linux just seems like a natural fit."
Avaya does this now, and they are porting more and more of there application VoIP services to LINUX, as well as Win2K like you said.
but a lot of solaris/sparc machines doing our telephony. The uptime is just too good on a *nix vs 2K.
errr....umm...*whooosh* *whoosh* Is this thing on ?
This is Mark Spencer's most recent project. Same guy that did Cheops and started GAIM. Really cool stuff.
Within the last 6 months I went through a phone system evaluation process. I was focused on IP telephony to a certain degree, so it was limited.
I agree that most items are being ported to Windows (scares the heck out of me, it's one thing for your web server to be down 6 hours, try having your phone system down for 6 hours).
The primary area where new development was being done, that wasn't Windows, I found to be in VxWorks. This makes sense to me since a RTOS really is a better platform, and at the same time, bypasses all of the Windows worms, etc.
Ok, I give up, why you?
Digium. A GNU/Linux telphony company based in Huntsville, AL. They sell T1 PCI cards for GNU/Linux machines and distribute a free as in GPLed software PBX. Check them out!
Disclosure: No, I don't work for them, but I have had lunch with them and they're pretty nice guys!
A company called West Telemarketing is working toward moving over their VRUs (Voice Recognition Units) from SCO to Linux by integrating the Dialogic (Intel) drivers into the kernel.
From what I have heard, things are in Beta but very stable and soon to be moving forward to production systems.
-- Knowing too much can get you killed, but knowing who knows too much can make you rich.
I used to work for Lucent/AG Communication Systems. The project I was on, their ClientCare call center system (think big... an entire in and outbound call center solution for arbitrarily large companies), ran on Solaris and FreeBSD. We had Solaris for the big Oracle Parallel server DB and FreeBSD tied the little bits and pieces together such as the CSR clients [which ran on Windows], the ISDN line management, and the playback of our utterly annoying hold music. It worked rather well, in the end. I think they're still doing it that way.
Here's a link to the product itself: http://www.agcs.com/productsv2/CallCenter/works.ht m
#44
VOCAL
Now I hope and pray that I will But today I am still, just a bill
The real bear in getting this to work was finding a modem suitable for use with vgetty; vgetty's docs list some voice modems known to work, but most of these are 5+ years old and $300 and up, if you can even find them for sale.
Clued in by a Usenet post, I found a modern modem that works: the 3Com 2976 Voice/Fax/Data modem. It sells in online stores for around $50. (Note that not all modems which purport to have voice functionality are supported, and controllerless "winmodems" are not likely to work.)
I also tried using Asterisk, but it wasn't really suitable for my voicemail needs. As I recall it did not handle disconnect detection very well, potentially leaving the phone off the hook for a long time. There was also a pronounced lack of any HOWTOs or detailed documentation available either with the program, with the PBX card I purchased from them to run the program, or on the Internet in general.
My sense is that Asterisk's creator actively discourages freely available documentation, in order to have people avail themselves of paid support. To his credit you do get one month of free support for the software and the card when you purchase the latter, and he was helpful in IRC when I spoke with him.
I do! But guess what? I didn't have to write the drivers because someone
already wrote them.
In my opinion, there is not a device driver problem here. Intel/Dialogic isn't the only vendor supporting Linux. And they don't support it out of the kindness of their heart: they support it because doing so helps sell hardware.
Quicknet has a low - cost 1 port card that will do the trick with Linux and Windows drivers:
http://www.quicknet.com
Also check out Pika for 4 port cards with traditional analogue and VoIP capabilities with Windows and Linux drivers:
http://www.pikatech.com
Aslo check out the Bayonne project. Linux based Open Source telephony system with interfaces to Quicknet, Pika, and other cards:
http://bayonne.sourceforge.net/
First and foremost I'm a wireless guy, landline is pretty much a black hole to me these days....
Telecom has been undergoing many changes at the lowest levels for a few years. Most UNIX systems in telephony are used as SCP's (Signaling / Service control points) / HLR's VLR's.. etc. A SCP will provide a service such as SMS, E-911, prepay, or something over the SS7 network. The SS7 network is at the lowest levels very similar to DAP, being a heavyweight protocol that requires its own circuits (ISDN, T1, ATM, etc.). While SS7 has been fabulous for creation of large and wonderful telecom networks it is becoming harder and harder to find people who understand even the basics of it. What's worse is the SS7 solutions of yesteryear (produced by say Lucent, NewNet, and Tandem) are no more. The newer SS7 solutions (say SignalWare, Distributed7, etc.) haven't really been able to cut the mustard. Things have been getting worse for a while, and people know it... but the fine people at ANSI and IEEE, Lucent, Nortel, IBM, and the like have come up with a solution. Make SS7 lightweight (I.e. IP based like LDAP).
Many things have happened in order to get SS7 (a very demanding protocol indeed) to work over IP. The first milestone was essentially dumping TCP for SCTP/IP. Much has been going on in this realm, the lk-sctp project has been busily cranking out code for the 2.5 series kernel, and will likely make Linux one of the first *NIX based operating systems to have a NATIVE SCTP implementation. Adding SS7 to the top of this is about as easy as creating an SCTP daemon.
While SCTP and the Sigtran suite of protocols (M3UA / SUA ) are moving ahead quickly there are other projects that are working on implementing a heavyweight implementation of SS7 - such as openss7, and even the PBX / softswitch project asterisk.
While all this may be nice and good, it may be worth noting that Inet Inc. has an SS7 network monitoring solution called GeoProbe. While some parts of the system run on a solaris server the actual cardcages and "proprietary" equipment actually run Linux. (at over 300k a site, that's a pretty big win for Linux).
As always I'd love to hear what's going on in other sectors of telecom with Linux.
I've worked for a few telephony companies in my time. . .two that stick out succinctly in regards to Linux are Priority Call (makers of Oryx) and Boston Communications (do the pre-paid calling back-end to most large cell carriers).
Priority Call (http://www.prioritycall.com) makes an uber-trunk-switch that does all types of cool stuff from one-number-anywhere to pre-paid calling to massive e-mailed voice-mail to web voicemail -- it is the swiss army knife of telephony (thier main competitor, BTW is Comverse). Priority Call has sold switches to the likes of Bell Atlantic and a lot of Mexican and South American cell vendors).
Anyway, their systems were passive-backplane PIIs (at the time) with Dialogic (owned by Intel) ISA-bus switch boards (Dialogic boards also have thier own bus to interconnect them to aide the ISA bus) and Mylex RAID controllers runnning 9 Seagate Cheetahs. They used SCO OpenServer because SCO was just about the only telephony OS 'back in the day' and it was pretty stable. As a side note they also legally used a fair amount of OSS in Oryx, including a hacked-up apache and ncurses.
SCO OpenServer, though, has not been actively developed on for a *long* time and not only does it show its age, but frankly it is just about the worst OS I have ever worked on (I don't mean to flame you OpenServer-lovers out there). Support was a bitch. Bug-fixing was a bitch becuase SCO was not longer developing OpenServer not to mention that later versions of OpenServer were hacks to old ones in attempt to add new features without the proper architecture. As much as I want to flame I'll leave my beef with OpenServer at that.
Needless to say the limitations of OpenServer were apparent and they found that *it does not scale* well at all. Thus, they moves their home-brew proprietary Oryx database to Tru64 using rack-mounted Alphaserver DS10s and kept OpenServer for the fron-end and switching to keep migration smooth (Comverse, BTW, uses Tru64 on Alphas -- which this whole push by HP to move to HPUX is going to really piss off a lot of telephony companies). For massive installs they used Sun Netra T-1s in a customer-specific manner.
Later, they finally realized that not only did OpenServer SUCK, but it was *expensive* too ($500 a copy). Thus they started to port to Linux and wrapped it into one massive migration strategy that included new hardware (Compact-PCI).
The fact of the matter is that people hate change. People complained about how Linux companies weren't doing so well not to mention that run-of-the-mill support people FEAR UNIX and the migration from the OpenServer database to the Tru64 was painful (had to re-do all the flow-chart-like step-by-step hold-my-hand this-is-how-to-use-unix cutsheets for some people)
My manager at the time (and the best manager I have ever had) sold Linux -- simply stated, "who cares if the Linux companies go under -- what is better security then HAVING THE SOURCE CODE TO THE WHOLE OS!). During the port, though, Linux had a few limitations that slowed the deployment:
1) OpenServer is such a hacked beast that porting to Linux from it was non-trivial.
2) Dialogic (the heart of the telecomm industry) did not make Linux drivers at the time. Thus they decided to move to NMS (Dialogic competitor) cards
that did support Linux as well as Compact PCI.
3) At the time Linux did not support hot-plug PCI which was one of the design specs and the main reason for moving to Compact PCI.
4) Not even NMS would ship source-code drivers -- only compiled modules. THIS IS A BIG THING as one can only run stock RedHat kernels or specific versions they support or else you'll get unresolved symbols or flakiness in the drivers. Face it, the stock RedHat kernel is *not* meant for telephony. Not only that, but the whole security argument of having the source code to the OS is negated because if NMS for some odd reason decided to stop developing Linux drivers then the company would be stuck with one version of Linux forever.
In the end it was not Linux's limitations that killed the migration but the fact that they rolled the whole migration into the massive hardware/software roll-over and when hard economic times hit and the person who spear-headed the project left, those that hate change won and the whole project was scrapped (some people think it is better to live with what you know versus venturing into the unknown, right?)
In summary, the things that I think would help adotion of Linux in the telecomm world are:
1) Above all else, open-source NMS or Dialogic drivers. People fear Linux companyies instability too much and if their vendor decided to stop supporting Linux it would screw them.
-OR-
2) A company come about that makes hard-core telecomm-grade switch boards with open source drivers that gives Dialogic a run for its money. I'm not talking about the "internet phone jack" guys, I'm talking about boards that can handle dozens on trunks (read T-1s). Dialogic used to be the main reason for companies not adopting Linux because they basically own the PC-based telephony market and they used to ONLY speak NT and SCO and trust me, as much as I hate to say it NT is better then OpenServer from a support and development point of view (although OpenServer is more stable then NT).
3) Keep moving forward with Linux on the desktop. Most people to this day *fear* UNIX and if Linux can be made common and user-friedly the managment types (and support types) that fear change will be less reluctant to let the engineers use Linux. It sounds convoluted, but this is how MS did it. Linux on the desktop indirectly helps all those who want to convince managment to use Linux a LOT as it shows that support costs will not be as high
as it is 'user-friendly' and they can hire monkey support cheap.
4) Linux clustering. Linux NEEDS good high-availibility open-source clustering. No matter how good your hardware is you can not get the telecomm "five nines" of uptime with one computer! A good first move would be a good filesystem that supports mutiple hosts sharing one fibre channel array.
Why do telephony companies migrate to NT/2000?
1) Tru64 is dead thanks to HP.
2) People are starting to fear that Solaris will go the way of Tru64 and future migrations are *very* expensive.
3) People fear UNIX and support costs are high due to this fear (need more geeky support people).
4) Dialogic only used to speak OpenServer and NT (I don't know if it is the same any more). NT is by far the lesser of the two evils in development and support (not reliability).
4) Managment fears Linux companies instability because they are thinking in the 'old school' support issue -- if a vendor goes under and you can't buy support your company is screwed. Please, educate them that HAVING THE SOURCE CODE TO THE WHOLE OS is teh best security. And please coerce NMS or Dialogic to make open-source drivers as their proprietary drivers negate the last argument!!!
As for Boston Communications, I did support for them and they used NT. That was one of the worst nightmares I have ever experienced. Try remotely managing hundreds of telecomm nodes all over the country over 56K frame-relay links using Remotely Possible (PC Anywhere clone). Not to mention the BSODs and managment blaming you when they could not report "five nines" to the carriers and thus had to pay them mucho $$.