Voicing my IPinion

Sending real-time voice traffic over a packet-switched network has far exceeded the "fad" stage. Most of the leading networking and/or telephony equipment companies now have generally available Voice over Internet Protocol (VoIP) products, including switches, gateways, and phones (both "hard" and "soft"). But many of the questions do still remain: What are reasonable Quality of Service (QoS) expectations? How reliable is it? Is it scalable? Will it integrate well into my existing circuit-switched environment?

An attempt to answer these questions, and more, follows in this, often opinionated, but hopefully informative, "IPinion Editorial." These opinions are the result of the author’s various beta tests, trials, and tribulations at the University of Maryland's flagship campus in College Park, MD, and do not necessarily reflect the collective opinion of the University, nor do they offer insight into each and every VoIP product and service available.

Like other leading research institutions and large, national and multinational corporations, the University of Maryland has been experimenting with, and utilizing, VoIP technology for some time. First and foremost, as an educational institution and community leader, the University wishes to learn as much as possible about emerging technologies. (Of course, VoIP is really beyond the state of "emerging.") More so, we are considering the possibility of having a single communications (voice and data) infrastructure, at least in new and/or remote locations; of taking advantage of an "infinitely" more extensive data infrastructure (allowing us to "connect" satellite locations that are beyond the range of our traditional voice network without having to install remote nodes / EPNs); and exploring the potential for more easily data-centric applications (CTI, Internet call centers, and so on).

On the (often bumpy) road to learning about, and understanding, VoIP, the University has experimented with, and put in place, a number of VoIP products including Lucent Technologies’ MultiMedia Communications eXchange (MMCX), wireless Palm Pilot voice cradles, Definity IP, and iCosm Collaborative Video (iCV); Cisco Systems’ Selsius products and Architecture for Voice, Video, and Integrated Data (AVVID); and 3Com Corporation’s Network-Based eXchange (NBX) products.

What have we learned? Let’s just say that if you’re trying to decide on a long-term VoIP strategy, I don’t envy you. The ramifications of what will happen if you do begin to deploy this technology are almost as scary as if you don’t. First, you must realize that there are three pretty distinct approaches to deploying VoIP technology: (1) Don't deploy at all, at least not now (I’d hate to be you in five years); (2) Forklift upgrade/replace all existing circuit-switched equipment with the new stuff (You’d better hope I have awfully good news with respect to the aforementioned QoS, reliability, and scalability issues); and (3) the hybrid approach (targeted at established enterprises, including most of us, that have significant investments in installed circuit-switched equipment, i.e., PBXs).

No matter what your approach, the challenges are numerous. Supporting the QoS that widespread voice traffic demands requires either some fairly significant over-engineering, or a darned reliable method of prioritizing voice packets and/or reserving a minimal amount of bandwidth. This is not a vendor-specific problem, but rather an industry-wide challenge that is being addressed in the "labs" and, more so, by the standards committees. For example, when the IETF Signaling Transport (SIGTRAN) committee completes its work on IPS7 (a proposed signaling layer protocol), the delivery of VoIP, as well as multimedia services, will have a QoS similar, if not equal, to that now associated with Signaling System 7 (SS7), the PSTN signaling network (or so they claim). The IPS7 standard also promises to support internetworking between SS7 and IP networks.

Reliability challenges are also fairly industry-wide, but the news on this front also seems good. The big vendors, such as Cisco, Lucent (now Avaya), 3Com, and others, are engineering their solutions to have the ability to power hard VoIP phones directly from the switch (provided you haven’t deployed a bunch of wonderfully fast, but non-conductive, fiber optic cabling everywhere). That, combined with redundant servers and other hardware components, will hopefully help eliminate "reliability" from the list of issues. Of course, the fact that most of these systems are NT-based might make some (me included) a bit skeptical. And what about the rest of the data network gear (hubs, routers, switches, and so on) and "soft" phones on desktop computers, not typically connected to some sort of fault-tolerant power system? This is another industry-wide, even Internet-wide, challenge that must be, and certainly will be, addressed over time.

Scalability issues are a bit more complicated, as they depend largely on your installed base of legacy equipment. Certainly, for those of you with a large amount of Lucent/Avaya Definity equipment, going with their Definity IP products (at least for now), likely makes the most sense. These products integrate very nicely with existing Definity PBXs, including almost identical feature sets and administration screens, and total scalability. Cisco’s Selsius products (especially the newly released vintage) might make a lot of sense for those (perhaps new) large enterprises with little or no installed legacy equipment. It seems particularly well suited for building a converged network "from the ground up." (Cisco's solution also features "Admission Control" which somewhat addresses QoS by selectively routing calls over either the WAN or PSTN, depending on WAN bandwidth availability at call-placement time.) For smaller enterprises, 3Com (who purchased NBX, one of the first IP-PBX vendors on the market) may have the most "tried and tested" line of VoIP products out there, but they do not, at least at the time of this writing, appear to be scalable in the "large enterprise" sense.

What other issues are there? Perhaps some of the most significant of them all are the economic arguments for, and challenges associated with, VoIP. There is the highly-touted potential for long-distance savings by utilizing IP trunk products (gateways) to "connect" multiple IP (or IP-enabled) switches over the public Internet. Certainly, depending on the need/traffic, the cost savings could be phenomenal, but let's be real for a moment. Not only is call quality, at least for now, far less predictable, but it would seem doubtful that large, national and multinational companies, who are already paying a measly three or so cents per minute for long-distance, are going to realize any significant savings, especially after all the high-tech equipment administration, costs, maintenance, and (more frequent) upgrades. Additionally, there are significant economic modeling challenges associated with VoIP. With new and different methods of service (potentially "selectable" bandwidth/QoS, packet priority levels, and other, possibly dynamic, parameters) come a whole slew of billing challenges. The Internet2 Consortium, with which the University is involved, is attempting to explore, identify, study, and address these issues. (Internet2 provides a wonderfully "clean" test bed for VoIP applications/implementations and subsequent data gathering.)

It would seem that the real cost-savings come, not from "free" calls over the Internet, but rather from the potential to wire an entire enterprise (or at least new construction) with just one set of cables for both voice and data. Savings accrue immediately through simplified installation and reduced wiring closet load. And truly integrated voice/data network management also holds long-term cost-savings through joint administration of user connectivity.

Which vendors will survive? There are good reasons the University has chosen to work with the ones it has. Lucent/Avaya, in addition to being a "partner" with the University on everything from technology trials to mentoring programs for technologically-gifted students, is one of the (if not the) leading telephony and communication research experts (thanks mostly to its infamous research arm, Bell Labs). As you've undoubtedly seen on the television commercials, Cisco has certainly not been shy in boasting that their equipment makes up the vast majority of the public Internet's gear. (They are, in essence, the Internet's "traffic cops", which could give them an edge in implementing proprietary QoS methods, such as prioritizing the transport of Cisco VoIP packets, though emerging standards will hopefully eliminate the need/desire for them to do this.) And 3Com is certainly a leader and innovator with respect to packaging useful products and technologies in consumer-friendly, cost-effective, easy-to-use offerings. (Their hubs and NICs have long been touted as the best available for the money.) But, certainly, these are not the only players, as the VoIP market (and technology in general) is moving so quickly that it is difficult to predict what will happen next week, much less in the long term. Besides these three vendors, a host of independent companies are producing truly leading-edge VoIP products and services. One thing is certain; the number of new technologies and acquisitions will only increase, meaning that we will be faced with even more choices as time progresses.

So where do you go from here? At the very least, identify a likely implementation path from the aforementioned approaches, and invest (at least a small amount) in some appropriate VoIP equipment now. As the advantages of a converged network undoubtedly push for solutions to the large (but shrinking) list of challenges, this is the time to gain valuable experience with installing, integrating, and managing VoIP equipment. VoIP is clearly the future of communications, and the future waits for no one.