Open Source Telephony

The increasingly disruptive phenomenon in the telecom world known as Open Source Telephony once had the same hobbyist, home-brew �persona� as VoIP did back in 1996. Indeed, the origin of open source telephony, Jim Dixon�s Zapata Telephony Project, or �Zaptel� (named after Mexico�s equally disruptive guerilla leader, Gen. Emiliano Zapata), consisted mostly of Dixon cobbling together his own PC telephony interface cards and writing open hardware drivers for BSD Unix under the GNU General Public License (GPL).

Mark Spencer then came on the scene, ported the code to Linux, and drew upon Zaptel technology to develop Asterisk, the world�s first (and most popular) software-based, open source IP PBX (News - Alert). In 1999, while still a computer engineering student at Auburn University, Spencer founded Digium (www.digium.com) to make the simple plug-in PCI bus telephony interface boards necessary for his software and its users to communicate with the outside world.

Digium (News - Alert) fostered the development of the Asterisk Business Edition, a professional- grade version of Asterisk that can handle voice and data transport over IP, TDM, switched, and Ethernet architectures. Asterisk (News - Alert) was now palatable to small and medium-sized businesses (SMBs), since it could run on a PC and work with legacy PBXs (e.g., Lucent, Nortel, Siemens (News - Alert)), IVR, auto-attendants, next-generation gateways, media servers, and application servers. Asterisk itself has all the functions of a good IP PBX, including call control, voicemail, with support for such VoIP-related protocols as H.323, MGCP, Spencer�s own IAX (Inter-Asterisk eXchange), and SIP, so it can be used with inexpensive SIP phones. (Recently Digium received a $13.8 million investment by the venture capital firm Matrix Partners.)

Digium�s Senior Software Engineer, Kevin Fleming, says, �We have two significant recent announcements. First, we�re releasing the betas of Asterisk 1.4, incorporating 20 or more significant new functions, along with interoperability and performance improvements and lots of other things that the community wanted badly, such as a generic jitter buffer that improves call quality during network congestion, T.38 so that IP faxes can pass through the server, whisper paging, more language capabilities, variable length touch-tone support for IVR applications, and support for Jabber, Jingle, and GoogleTalk, to name just a few improvements. Some of the development work was done by members of the open source community, not Digium.�

�Our other big announcement is that we have a new piece of hardware, an embedded Asterisk appliance,� says Fleming. �It�s a little box that has modular analog ports, CPU, RAM, Flash memory, echo cancellation, Ethernet ports, and everything you�d want on an appliance. It runs Linux for an embedded platform and unmodified Asterisk 1.4 � but it can�t run Asterisk 1.2. This device will eventually be a retail item that Digium will sell, but right now it�s being released as a Developer�s Kit. It�s for people who have an interest in producing some sort of great vertical market PBX, or a gateway or some derivative of what Asterisk can do, but they don�t want to do it on a regular-sized PC server. Now they have a compact hardware platform on which to build the next great world-beating product.�

�It�s even smaller than a pizza box.� says Fleming. �It�s about 11 inches long, six inches wide, and slightly over an inch thick. It looks just like what you�d expect an eight-port analog terminal adapter to look like, except this runs Asterisk, so it can do anything Asterisk can do. It�s modular, so the base unit has no analog ports, but it uses the same one- and four-port analog modules that our existing PCI analog cards use. So you can populate it with up to eight ports of station or line ports, depending on your choice. You could use it just as a terminal adapter, and we envision that there will be customers, such as Vonage (News - Alert)-like operators, that may want to deliver a highly functional eight-port terminal adapter to a business customer, rather than sending eight individual one-port ATAs [Analog Terminal Adapters]. But the box could also be used as a fully functional PBX and it could even be used with no analog connectivity at all � you could purchase it without analog modules and just use an IP carrier and IP phones.�

�I�ve noticed over past six months or so that open source development philosophy is starting to permeate into the world of corporate development,� says Fleming. �That won�t necessarily affect a company like Microsoft (News - Alert), but if you look at internal development in large corporations, they�re starting to learn from the way the open source community develops things, involving as many people as possible and making releases as quickly as possible, and just being completely open about everything that you�re doing. It appears to be a much more efficient way to produce software than the traditional model. Large companies are slowly becoming more comfortable with outside open source communities producing software that they want to use.�

�For example, when a company asks if open source is something you can really rely on,� says Fleming, �we point out to them the huge number of Web servers running Apache � no one even thinks of it as �open source� any more. It�s just Apache, the Web server you use if you�re not using Microsoft. We would like to see Asterisk become a comparable tool in the telephony space, so if you�re not using a closed source IP telephony platform then the one you�re using is Asterisk. For Asterisk to happen to be open source is a useful thing, but it�s not necessarily the reason that you�ll either pick it or stay away from it.�

Picking Up the Asterisk Ball and Running with It

Once Asterisk began to catch on with businesses as a low-cost alternative to traditional, pricey PBXs, other companies began to build advanced hardware and software �scaffolding� around Asterisk, creating systems of great scalability and reliability.

Once such company is Sangoma (News - Alert) Technologies Corporation (www.sangoma.com), which offers a range of solutions and support for software-based PBX and IVR voice systems, their WANPIPE� internal routing solutions, API communications toolkits for OEM users, POS (Point-of-Sale) interface cards and their WAN EduKit for demonstrating the inner workings of WAN protocols such as ATM, Frame Relay, and X.25 in an educational setting. Their solutions encompass traditional legacy protocols as well as IPbased voice and data technologies.

Asterisk aficionados and developers will be familiar with Sangoma�s portfolio consisting of its popular Advanced Flexible Telecommunications (AFT) family of PCI-based cards with T1/E1, T3/E3, TDM, Analog voice and data, ADSL and serial interfaces that can help turn a PC server into a voice or data gateway.

Recently, Sangoma announced that it�s now shipping a complete product line of enhanced PCI Express (PCIe) bus cards, based on its ATF card family. Sangoma�s Octal-port A108, Quad-port A104 and dual-port A102 T1/E1 cards, as well as its A200 analog voice system, are now all available in PCI Express format and can exploit this new switch fabric architecture. The change to PCIe is completely transparent to the user and all of Sangoma�s drivers, utilities, and APIs continue to work exactly the same way for PCIe as for the older PCI standard busses. (Sangoma will continue to produce standard 3.3v/5v PCI versions of the A102, A104, A108 and A200 cards to support legacy PC servers.)

Sangoma�s President and CEO, David Mandelstam, says, �Asterisk is open source, there�s no question about it, but their business model is more of a �mySQL model� in that it�s open source, but they do sell it in a commercial version. Now, mySQL does very well with that model; it�s a closed system and it�s a large enough organization so they can do their own development and they don�t get much help from outsiders. I suspect that�s kind of what�s going to happen with Asterisk. Asterisk will be limited by what actually can be written by the developers at Digium and a few other people round-and-about who, for one reason or another, are happy to do the work and contribute so that Digium can make money out of it. But, in general, the innovation in this business will occur outside of Asterisk because there isn�t a compelling reason for people to give their code to Digium so they can make money.�

Mandelstam elaborates: �Take, for example, an open source application such as Trixbox � formerly [email protected], a free home or office PBX phone system based on Digium�s Asterisk PBX. You simply take the open source of Asterisk and meld it with these other applications. Trixbox is interesting because its developers can never sell it, because they don�t own it. Asterisk can never sell Trixbox software, even though they�re doing a lot of development, because they don�t own it either. So the product that Trixbox produces is true open source: it consists of pieces of code taken from all sorts of places, and it has been put together into a package that includes bits of code not in Asterisk.�

�Since Trixbox is open source,� says Mandelstam, �if they want to grab our WANPIPE drivers and put them in their own code, they�re perfectly welcome to do that, because our drivers are open source, as is UniCall�s �spandsp� library of DSP functions for telephony, and as is Asterisk, for that matter. Trixbox or someone like them may decide that Asterisk does an excellent job of conferencing, but the SIP stack could stand improvement. So, throw the stack away! There are better SIP stacks out there. Just grab one from somebody else, such as the SipExchange SIP stack, or the one from Yates, or the FreeSWITCH open source telephony application. The Digium model won�t let you do that.�

�Modular extensions to Asterisk take on an open source life of their own,� says Mandelstam. �With the Digium model, everything has to be written by Digium employees, who work for Digium. They may or may not be paid. And they probably aren�t doing it to gain notoriety, since their name gets taken off of the code. In a true open source model such as Linux, the contributions are made by all sorts of people freely and nobody really owns it. The code is protected only by the Free Software Foundation [www.fsf.org].�

Rustling Up Some Scalability and High Availability

For several years some industry observers (Yours Truly included) have suggested that some large company should step forward to support open source telephony the way IBM (News - Alert) supports Linux. Interestingly, the open source telephony industry appears to be moving in that direction anyway, mostly because of a multi-prong approach by many companies that are adding scalability, reliability, and security to Asterisk.

Take for example Ranch Networks (News - Alert) (www.ranchnetworks.com), which has developed network appliances specifically for Asterisk and, by doing so, has increased Asterisk�s scalability and security. They use a clustering scheme to achieve this, which makes their approach resemble a Web server farm. So, if there�s a failure on one of the Asterisk servers, the other servers will pick up the slack as load balancing occurs.

Moreover, Ranch Networks� little boxes are stuffed with more functions than you�d normally find in a whole room full of equipment: They support Security Zones, Bandwidth Management, VPN, VoIP, Load Balancing, Real-Time Server Health Monitoring, IP Multicasting, Usagebased Accounting, and Layer 2-4 Switching. Their equipment allows service providers and enterprises to control network resources on a per-call basis, including per-call security and per-call bandwidth allocation. It certainly lends credibility to their slogan, �Carrier class products at enterprise prices.�

Ram Ayyakad, founder and CEO Ranch Networks; says, �We are enhancing the reliability and scalability aspects of open source, specifically Asterisk. We contributed a certain portion of our efforts to the open source code, which would, in real time, allow a PBX to send policies to an external appliance. It�s a standard-based interface. So when a call is going through, there are a variety of policies that get exchanged on a protocol basis between the open source PBX, the Ranch appliance, and an external appliance.�

Ayyakad continues: �The Ranch appliance, when it receives the commands, enforces these policies on the appliance. The policies received from the open source PBX are placed into several categories, some of which are based on security, some for call quality aspects of the PBX, some are related to the NAT traversal aspects of the PBX, and some relate to improved scalability, and whether we do RTP bridging, or whether we do load balancing and things of that sort.�

�Our contribution to open source is to add to it what we are good at,� says Ayyakad. �Open source is often viewed as a small enterprise play. We come with a data security background and on top of that we know VoIP-related security, so we add VoIP security as a typical component of open source PBX deployments. Since we have a background in being able to terminate both data and voice, we add a converged appliance to the solution so that an administrator can terminate both data and VoIP and feel confident that a voice call will to go through, regardless of the total data traffic that�s flowing at any given point.�

�We have added 1 + 1 high availability to the system,� says Ayyakad, �so an admin can configure both an active PBX and a standby PBX. We monitor in real time at the applications level what�s happening and the health of the PBX. So, if at any time the PBX is not up to the mark for delivering phone calls, we can, in real time, switch over to the standby PBX.�

�In September we added what we call a clustering solution, based on our VoIP Matrix technology,� says Ayyakad. �This allows a user to configure one or more Asterisk servers as a cluster. We can monitor each and every element of the cluster in real time, the same way we do monitoring in the case of a high availability solution. The user has the ability to add or remove an element [an open source PBX] in the cluster.�

�The upper bound of our system is currently 2,000 simultaneous calls,� says Ayyakad. �We have plans in the first quarter of 2007 to scale up to 50,000 simultaneous calls. In the next 6 to 12 months it will not be surprising to see about 10,000 simultaneous calls at least. The interest in being able to processes a large number of calls is mind boggling, and so you�ll see fairly decent-sized installations when we introduce even larger scale support on our appliances.�

Features, Features and More Features

One favorite theme concerning open source telephony is enabling a struggling smaller business to have the same call control and other telephonic capabilities as a larger, prosperous enterprise.

Chris Lyman, founder and CEO of Fonality (News - Alert) (www.fonality.com) says, �Most Asterisk companies are adding a more friendly and sophisticated front end to the underlying Asterisk code. We took a more holistic approach. We reworked things from the back all the way to the front. We started at the hardware level where it interfaces with the phones, worked our way up through the LAN, re-wrote all of that code, came up into Asterisk, stabilized it and extended the feature set, then brought it all the way out to the application layer. In short, we�ve written more lines of code in and around Asterisk than Asterisk has to begin with.�

�Thus, our solution goes beyond the ten or so companies out there that have done little Web front ends for Asterisk,� says Lyman. �Still, we�re GPL-compliant. When we make changes to Asterisk, we follow �the law� and release our source code, but we�re a significant fork from Asterisk, both in terms of reliability and the feature set.� �Our PBXtra product has two different levels,� says Lyman. �It extends down to the Asterisk feature set, and we do that where nobody really owns the intellectual property. But when it extends up into the application layer, which is a significant piece of it, we retain the assets to it, because it runs as a separate program.�

�Our market is primarily North America,� says Lyman. �We have 1,000 deployments in 22 countries, but 90 percent of that is in the U.S. The product has been taken overseas mainly by resellers.�

�There is no theoretical scale limit to our product,� says Lyman �However, we tend to focus on the 300-seat-and-below market, because the space above that is very competitive, with companies such as Avaya (News - Alert) and Cisco selling there. About 60 percent of the American workforce works in businesses of 200 seats or fewer. Many people don�t realize that small to medium-sized businesses (SMBs) comprise most of the American economy. And it�s that segment that we think is getting ripped off.�

�Our key product is PBXtra,� says Lyman. �It�s a truly enterprise-class PBX, but it sells for just $3,000 for 10 phones. Its base package is $995, which consists of our server and our software that gives you IVR, voicemail, conferencing, and PBX replacement functions. Then you start adding phones. We don�t even make the phones; you can use phones from Polycom (News - Alert), Cisco, Aastra, and so forth. Our call center edition of the product, which is very sophisticated, is only $1,995. It gives you the base features plus a true queue system, and the ability to build distributed call centers around the world. We�ve even extended the product�s capabilities so that agents can participate inside of a call queue from their cell phones.�

�Small companies are looking for a product that�s priced appropriately, and easy to use,� says Lyman. �They don�t really care if it�s open source or closed or partial, or whatever. They just want it to work, since they probably don�t have a big IT staff on the one hand or an $80,000-a-year in-house Linux expert on the other. That�s the key thing that Fonality has done inside the Asterisk space � we�re commercialized or productized the very complex, unwieldy Asterisk platform. Small businesses come to us because they don�t have that kind of in-house help, but they still want the low-cost telephony product.�

Mobile Open Source

Thanks to its inherent �streamlined� nature (along with a lack of commoditization, flexible application capabilities, and low cost), Linux has been embraced by mobile handset vendors. Nearly all leading and emerging telecom equipment makers have developed and are shipping Linux-based mobile devices, including Datang, e28, Haier, Huawei (News - Alert), Motorola, NEC, Panasonic, Samsung, and ZTE. In 2005, handset OEMs in Asia shipped nearly 15 million phones with more than 20 phone models running Linux. Of that, Motorola (News - Alert) shipped more than 5 million handsets on MontaVista Linux.

At MontaVista (www.mvista.com) the Director of Product Marketing for Mobile and Wireless, Paxton Cooper, says, �If you look at the mobile phone market over the last four to six years, the trend has been to jam in as much multimedia complexity as possible in terms of high-end gaming applications, messaging, camera phones, and mobile TV. Five or six years ago, all of the complexity in mobile phones was centered on communications and how you actually went about making a GPRS, CDMA, or Wideband CDMA phone call. Getting things to work in the old days was the big thing. But then, it increasingly became a question of how you could manage all of the software complexity associated with the multimedia requirements that the operators, to a certain extent, have jammed down the throats of handset manufacturers. Everyone is piling feature upon feature into the phones. Millions of lines of code have been incorporated into these devices as they have moved from just dealing with pure communications processing to application processing, where you now have multimedia-rich and visually rich devices with all sorts of bells and whistles to lure subscribers into buying and using these devices, as well as to help manufacturers differentiate their devices from all of the other handsets on the market.�

�This spike in application complexity in handsets has presented manufacturers with several problems,� says Cooper. �First, how do they manage their cost in terms of having to license all of these third-party applications and such to incorporate into their device to meet the requirements which the operators have specified? Secondly, how do you build, install or otherwise have a platform in place that can continue to meet these rigorous multimedia requirements? Most mobile phones are powered by real-time operating systems [RTOS] such as OSE, pSOS, Nucleus, and so forth. These are sort of home-grown or licensed, proprietary real-time operating systems that were perfectly sufficient for voice applications but are beginning to run out of steam when you decide to support all of the feature-rich multimedia applications and requirements driven into these devices.�

�That�s what really opened the door for Linux to become a solution for these handset manufacturers, as they go about developing their next-generation platforms,� says Cooper. �Linux is a robust, modular, high-powered, feature-laden operating system that�s supported on most of the new hardware and baseband platforms these manufacturers use for their devices. It�s a logical fit and a next step for these manufacturers as they being to develop more and more capable phones.�

�Half of these manufacturers are trying to differentiate their devices and also control their costs,� says Cooper. �They�re very reluctant to be beholden to Microsoft and use their Windows CE platform for their devices, and so they�re looking for alternatives.�

�Those are really the main drivers that we�ve seen and it�s why manufacturers have reached out to us and are looking to either investigate or go ahead and adopt our product, which is mobile Linux, into their phones,� says Cooper. �Again, just to recap, it�s really this need to respond with a much more capable platform, and respond to the multimedia complexity and requirements that are being driven into modern handsets and phones. It�s the need to have flexi bility with the platform in terms of being able to create a differentiated phone that doesn�t look like a Microsoft or Symbian handset and that actually has some uniqueness to it. It�s also just the technical arguments for Linux in terms of it being a platform that many engineers are familiar with and comfortable in developing for a very modular platform. It�s a more modern operating system that�s compatible out-of-the-box with the hardware that they�re using to build these devices. And Linux also supports all of the key technologies that they�re looking to incorporate into these new devices.�

Safety First

�Additionally, there�s a misconception that because it�s open source software, it�s somehow not secure,� says MontaVista�s Paxton Cooper. �When you talk about the most insecure operating system out there, it�s clearly the Microsoft platform, and ironically, that�s as proprietary as it gets. You have to differentiate what it means to use an open source platform from the fact that the source code is broadly available but, at the same time, separate that from the actual implementation � of how it�s actually deployed into your platform and device.�

Converging Operating Systems?

One wonders if such mobile operating systems will grow in size and complexity until they begin to resemble the desktop. Could both the desktop and mobile worlds converge to a single operating system? Under IMS, services are supposed to be able to roam anyway. Could users in the distant future end up relying upon a single operating system for both desktop and mobile devices?

�I don�t think so,� responds Cooper. �But I think the mobile device experience will resemble what you�re experience is on the desktop. If you look at the compute capabilities of the mobile phones in our pockets today, they far outstrip the computing capabilities of desktop systems from seven or eight years ago. You�ll continue to see more capable operating systems in these devices and they�ll begin to closely resemble what you have running on desktops, but you have to remember that the requirements of these phones are still very different from what you have in your desktop systems.�

Handholding and Batteries Included

Although open source telephony is perhaps the world�s most egalitarian approach to providing an IP PBX to any organization, not every company has a well-informed IT staff, let alone a resident Linux expert. There�s still a place in the open source telephony industry for such time-honored folk as integrators and professional services people.

Take SIPBox (www.sipbox.net), which designs, implements, and supports end-to-end telephony solutions for medium to large enterprises (200+ users). Specializing in Asterisk, SIPBox offers a complete solution, starting with on-site network design and integration and continuing on to management and 24/7 system maintenance.

Chad Agate, Co-Founder and CEO of SIPBox, says, �We started our company in 2000 as The Cipher Group, doing just general IT consulting, systems integration, and systems design. We implemented Cisco�s Call Manager, Unity, that kind of stuff.�

�Last year we received training and were certified on the Asterisk platform,� says Agate. �We decided that open source telephony was ready for prime time. So, as of January 1, 2006, we reincorporated under the SIPBox name, focusing solely on deploying Asteriskbased solutions. We can provide an endto- end solution from the network readiness assessment all the way through implementation and support afterwards.�

�In the purest form, we�re an Asterisk integrator,� says Agate. �We utilize the Asterisk Business Edition right out of the box. We come in and do the network readiness assessments, make sure that the network can handle VoIP, make configuration changes to the network that are necessary, and implement the Asterisk servers on the network. After we do the handholding and get the system up and running, we then support everything afterwards.�

�We like to deal with companies with at least 200 endpoints and up,� says Agate. �Asterisk can handle that right out of the box. You simply have to scale the servers properly, and design the Asterisk infrastructure correctly. Indeed, we have sites of 250 and 300 endpoints.�

�Economics initially drove the popularity of Asterisk, but now it�s also the feature set that has become available in its latest version,� says Agate. �Organizations may look at other solutions, but when they see the numbers and the features, they go with Asterisk. We�re having our greatest success in education, municipal and state government, and financial services. At the moment, our entry into many of these companies is via the voicemail application. They may have an Octel system that is reaching the end of its working life, and they need a new messaging platform. They don�t want to spend the money for a commercial system, and so they�re looking at Asterisk to provide voicemail services. You could say that voicemail is our �foot in the door�.� IT

Editor�s Note: A larger version of this article with more extensive interviews can be found among �The Zippy Files� online at http://www.tmcnet.com/tmcnet/columnists/

Richard Grigonis is Executive Editor of TMC�s IP Communications Group.

If you are interested in purchasing reprints of this article (in either print or PDF format), please visit Reprint Management Services online at www.reprintbuyer.com or contact a representative via e-mail at [email protected] or by phone at 800-290-5460.