Every one of my CIO clients has told me that VoIP deployments are on their list of projects to implement. They also tell me that this market segment is confusing, making them nervous and causing them to stall implementation plans. Compounding their confusion is that they perceive a lack of a clear architectural roadmap to follow. They´re right! When you look at how VoIP products are deployed, you invariably find a number of gateways to connect into the PSTN, PBX, phones (SIP, MGCP, H.323 or analog and digital) and the Internet. At every gateway there is a black hole that segments one part of the voice network from another, making management and administration a black art at best.

It´s no wonder that, after 4 years, the VoIP market is still largely relegated to the low end of the enterprise market. Since the very beginning, the design center for the first generation VoIP offerings has been small workgroups and branch offices. These VoIP systems from 3Com, Cisco, Avaya, Nortel, Shoreline, et al have proven to replicate basic circuit switched voice service at a lower cost. In the past few years, multi-site networking of VoIP solutions has been proven to reduce cost through toll bypass. First generation VoIP solutions´ key value proposition is improved user productivity through unified messaging and reduced total cost of
ownership through convergence. But through it all these solutions are still largely terminal-to-host based voice systems. So how will these voice networks change and evolve. I offer a road map.

First Generation Voice Systems: Terminal-to-Host

Circuit switched or TDM-based PBX voice systems are, in essence, a terminal-to-host system. Analog and digital phones are dumb devices, which are slaved off of a centralized host, the PBX. The PBX hosts 90% of the proprietary software and distributes features and functionality down to the phone terminals. A hub and spoke physical network connects the dumb phones to its host. Hosts interact with and are connected through the PSTN or private lines. The circuit switched voice network is very similar to IBM´s old SNA/mainframe network architecture, and the similarities go beyond a common terminal-to-host networking approach.
The PBX industry structure, like SNA/Mainframe networking, is proprietary and controlled by a few companies. SNA/Mainframe networking was administrated and managed centrally. Purchasing was centralized. Computing was centralized. The result of all this centralization was that new applications took too long to develop, cost too much, and often missed their productivity enhancement mark. This gave MIS, the controllers of the glass house, a reputation for being out of touch with departmental managers.

Moore´s Law and open systems worked its magic on PCs, IP, workstations and PC LANs to give way to new approaches of delivering enterprise IT services. The PBX market will change in a similar way to the SNA/mainframe market. There are two key areas of change. First, voice will be transported over IP just as SNA is today. Second, VoIP solutions are moving toward an application environment that is quicker to deploy, is more flexible to meet business requirements and is open to many vendors
and/or independent software vendors.

PBX Software Feature Set Gets Chiseled Away

PBX software is a closed environment meaning that changes are slow to occur, expensive to make and controlled by the PBX vendor. Most PBXs are made up of some 4+ million lines of code. This code delivers reliability and a 670 +/- voice system feature set. While the feature set of PBXs is huge and contains years of development work and intellectual property, understanding from whence it came divulges it´s future. PBX software was built over the years by adding a custom, one-off feature, done for a single customer to win a deal. In short, many customers said to their vendors ?¬¢‚Äö?ᬮ??¨if you build this feature into your product, we will buy your PBX?¬¢‚Äö?ᬮ¬¨??, and Lucent, Nortel, Mitel, and all the other PBX vendors followed suit to win business.

A section of PBX code is reporting-related and a lot of that functionality can now either be handled by native Windows functions, or can be quickly obtained and customized with Perl/SQL/Visual Basic scripts, etc. Plus, the market now wants these reports to be viewable in a browser, so the old reporting formats are even less appealing. For those non-reporting features, its unclear how useful they may now be to a broad market since they were built to support an enterprise´s application that may have been
retired, or the single customer is no longer interested in the feature. Much of the massive PBX code base was designed to bring PSTN features into the PBX. Since many of these features have been pulled into IP PBXs in significantly different ways, the “old” code base, over time, runs the risk of becoming less important. For example, pushing Caller-ID from the PSTN down to a desktop phone happens differently in the IP world than in a switched PBX, making a chunk of the PBX code increasingly irrelevant. Another very large block of this code supports inbound/outbound call center features, which should enjoy a long shelf life.

Cisco has tried to replicate call center features within Avvid, with its 1999 acquisition of GeoTel, but by all accounts has failed.
Many other non-reporting features can be added quickly by a capable C++ system integrator / enterprise programmer using open APIs. In fact, the XML programming capabilities in many IP PBXs probably exceed what could be done with any TDM PBX architecture. For example, in many cases, the old S.100, CTI-style APIs in PBXs will need to be replaced by web/intranet-oriented APIs. In essence, what is happening to PBX software is what happened to the mainframe: open systems are unlocking a ve ndor´s grip and opens up the market to many. Many PBX features will be replaced by new paradigms. For example, ?¬¢‚Äö?ᬮ??¨Camp On?¬¢‚Äö?ᬮ¬¨?? gets moved from a star-code on the phone to a mouse click on your PC — maybe even a mouse click in your Microsoft
Messenger “buddy list.” As you start to look at the PBX feature list, you realize that much of the old features can be done differently and better in the VoIP domain.

A Bottoms-Up Market Approach

The transition from mainframe to client-server was a bottoms up market adoption. Most IT markets start with early adopters willing to accept bugs in exchange for higher performance or a unique feature set. SNA and VoIP both share this bottoms up approach, with implementations starting in the low end, low tech market, then moving up to larger more complex installations. Novell, with the PC LAN movement, and Sun with Workstations, epitomized this bottoms up approach. Novell and Sun targeted the mainframe with a containment strategy, meaning that they would build workgroup networks and PC LANs that surround the mainframe and, in the end, choke it off from further investment. At the same time, networking companies were offering a wide range of alternatives for transporting SNA via data link switching or encapsulating SNA in IP, as well as channel attachment interfaces for routers. The combination of a new application environment and transport mechanism put big cracks in the IBM armor. This is very important. Novell, Sun and the networking companies understood that the mainframe/SNA environment was
administrated/managed and budgeted centrally and that they needed a strategy that would be the direct opposite. Sun went open systems, meaning IP and its associated APIs for application development. Novell´s NetWare went after departmental managers that had their own budget and authority. Networking companies went after an economic efficiency story that converged SNA, IP, IPX, AppleTalk, etc to win business. What departmental managers got was control of their computing environment to meet business demands that central MIS was simply too slow and expensive to address. The network managers got flexibility and cost
savings. VoIP is following this same bottoms up path to market.

What was key to Novell was its channel to market. It was its large and loyal network of VARs that reached departmental managers and created a large barrier of entry to its competitors. The VoIP market has to move in the same way. VARs can customize VoIP solutions to meet business objectives. Those VoIP vendors that choose to go to market with a reseller
strategy wi ll find themselves with dwindling market share. VoIP is going to be all about customizing select voice services to meet particular business requirements.

VoIP has found success in workgroups, small businesses and branch offices. VoIP installations will only grow over time and it too will contain PBX expenditures and feature expansion. The VoIP market is increasing being split between hardware and software vendors. The hardware vendors sell phones, IP PBXs, Ethernet switches and gateways. The software vendors are attacking the PBX code base with a new set of tools based on open systems.

Today´s VoIP market value proposition is largely rooted in transport and toll bypass cost savings plus productivity improvement through unified messaging. Both first and second-generation VoIP products largely offer more efficient and cost effective ways to deliver existing voice service rather than entirely new features. As the VoIP market enters into its 3rd generation of product, its value proposition will expand to include business driven features that add voice functionality to the computing experience. The fact is voice services are tightly linked to enterprise profit drivers, they always were and always will be. The first generation of VoIP products have demonstrated that they can deliver basic voice features faster, easier and less costly then in today´s PBX based systems.

Second Generation VoIP: A Bridge Between Old and New Worlds

We are now in the second generation of VoIP products, thanks to companies such as Avaya (with ECLIPS) and Nortel (with the CSE MX). These companies have put in place the key technology to bridge the gap between their legacy systems and the first generation VoIP solutions. The Avaya ECLIPS architecture and Nortel Succession products are very important since they are the glue that ties the TDM circuit switched network to the IP voice infrastructure. These products will allow a wide range of voice end points to communicate over an IP network to their respective PBX. This bridge will allow the PBX to take its place as a server in a 3rd generation of VoIP systems, which will be a distributed network of client-server voice services.

Third Generation Voice Systems: Distributed Client-Server Model

VoIP solutions need to expand and offer effective business tools from an unbundled set of features delivered from an open platform. This is what 3rd generation VoIP will deliver. First, VoIP is moving to a distributed client-server-computing model. This is true from an architectural, industry landscape and rate-of-adoption point-of-views. Second, 3rd generation VoIP architecture will be made up of smart clients, not dumb. SIP looks like it has won the voice client protocol battles over MGCP and H.323. In fact, 2003 may very well be the year of SIP. Third, application servers, some on your intranet some on your extranet, some on the internet, will deliver voice services such as v-mail, call management, unified messaging, CRM, directory, etc. A set of industry wide APIs needs to be agreed upon to let 3rd party application developers build new voice services into the VoIP environment and inject a huge amount of innovation and creativity. Connecting clients to voice application servers will be via a quality aware IP network made up of Ethernet switches and routers. There is much work to do on this last point and VoIP vendors will differentiate on it.

This distributed client-server architecture will streamline the gateway problem, put VoIP into a known model that many understand and unleash the creativity found in open markets. I believe that profit driven productivity enhancement will be found in the full integration of voice service and features into the networked computing environment. That computing environment includes the Internet, PCs, laptops, palms, SIP phones, PBXs, analog/digital phones, 802.11 phones and transport etc. In short, as the industry enters the 3rd generation of VoIP, it will be an inclusive model. With voice services located and distributed on servers and on the internet, VoIP will scale to enable a rich and varied user experience to meet a wide range of business objectives across all industry segments in the world economy.

Every major VoIP player has adopted this distributed client-server VoIP model and is moving toward it over the next few business cycles. Clearly the rate of adoption is varied among the vendors. Avaya, Nortel, 3Com, Mitel, Cisco, Shoreline, PingTel, and everyone else are building to this model. Sure, 3Com, Cisco, Shoreline and PingTel don´t have an installed base of PBX customers, and Avaya, Nortel and Mitel can, to a degree, regulate the transition from 2nd to 3rd generation VoIP solutions. One thing is for sure: Avaya and Nortel have embraced the move to VoIP more than IBM embraced the transition to client-server computing. Both have a financial interest in building a new IP infrastructure to support voice, a deep understanding of voice applications and large customer bases. One important note, PBXs are not going away, just like mainframes have not disappeared. Clearly development will shift away from the PBX, but there are features that are best delivered in a centralized model. One example, is the support of multiple end points that share, or are able to view, the status of other end points. For example, a secretary´s phone has an appearance of his/her manager´s phone line so that he/she can make intelligent messaging choices. In this scenario a feature affects multiple end points simultaneously. Making this work requires tight cooperation and may prove difficult to work in a distributed environment. So the PBX will have a place in 3rd generation VoIP solutions.

The above discussion does not mean that all 3rd generation VoIP systems will be the same. This new generation will have more differentiated architectures as the traditional players start leaving their old technology behind and cross the chasm with completely new approaches. The basis of competition and differentiation between 3rd generation VoIP architectures will be around scale, open standards and interoperability, ease of management, ease of use, increased reliability (without single points of failure common with PBX and server management) and designs that drive costs out of installation and management. A rich set of bundled and unbundled voice applications with the right hooks for application developers will further differentiate 3rd generation VoIP vendors. Also, those that are excellent and excel at rapidly customizing their VoIP solution to meet customer business requirements will find a warm and long lasting business relationship with their customers.

The road map of VoIP is, in essence, a transition from terminal-to-host to a distributed client-server model. Two bits of advice: we are in hard and difficult financial times, with many start-ups filing for Chapter 7 liquidation. As you look to build out your VoIP network, you will invariably include new companies in the mix. I would put cash flow positive and/or profitability and/or capitalization out to 2004 as a larger weight on my basis of competition criteria for this category of company. Second bit of advice, many companies will wave the flag of open VoIP systems but will attempt to lock you in to their proprietary solutions. Case in point Cisco´s in-line power, skinny SIP, Survivable Remote Site Telephony or (SRST) are proprietary protocols that seek to lock customers into Cisco infrastructure and create a barrier of entry for new comers. As we enter into this new generation of VoIP, we should not forget an important lesson learned last time we transitioned an IT industry segment to open systems, buyers beware of closed systems masquerading as open.

Posted 2 August 2002 at 6:45 pm under Lippis Report. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.