It's the spring of 1998. I'm having lunch with two old friends, David Tucker and Richard Platt, at an Outback Steakhouse restaurant. We are discussing the company they have cofounded, Selsius Systems. They tell me that they have developed the perfect technology for the integration of voice, data, and video. Thinking back on that lunch meeting now, I remember stirring my iced tea with a sense of déjà vu. My mind immediately wanders back to 1986.
In 1986, while working for ROLM Corporation, I was part of a team that sold and installed a voice and data integration solution to a large university in Dallas, Texas. David Tucker was the sales manager leading the team. We installed over a thousand digital telephones with data connectivity to allow the connection of asynchronous devices to the ROLM CBX selected by the university.
One of the key criteria for the university was their desire for a more cost-effective means of connecting data devices to various data sources, both internal and external. They looked at the PBX as the logical staging point for integrating voice and data. They bought into the promise of easier administration, of single wiring to the desktop, of shared access to various data hosts—all in the name of saving dollars.
Years later, in 1994, David and I were together again, this time at Intecom, where we continued to blur the lines between voice and data with a product called InteLAN. InteLAN was a connectivity hub integrated into the Intecom PBX and was the brainchild of an engineering team headed by Richard Platt.
Fast forward to the Outback Steakhouse in 1998. I am listening to David and Richard excitedly discuss this new product and where it is going to take the industry. I remember the single, simple thought that jumped into my head at that moment:
"Haven't we been down this road before?"
I suspect that many people, when they first hear of IP telephony (IPT), react in much the same manner.
"Here we go again."
"Voice and Data Integration, Part 2."
I can't blame people for thinking this, because in many aspects, it is true. Integrating voice and data into a single platform is not a new idea. (Some might argue it is not even a good idea, but I'll cover that later.) The PBX manufacturers championed this concept back in the 1980s. The idea at that time was to use the PBX and voice infrastructure as the focal point for integrating the two technologies. In many respects, this made perfect sense, primarily because of the high reliability perceptions of the PBX and voice infrastructure.
This chapter explains why a PBX, despite its high reliability, is not a solution for convergence. It also examines what makes IPT different from earlier approaches to convergence, and discusses application development as the key to successful IPT deployment.
The PBX as a Convergence Platform
The PBX is arguably the most reliable technology mankind has created and so it seems a logical choice to use as the platform for integration. If you talk to most people, the perception they have is that although their mainframe might hiccup and their network might snooze every now and then, the telephone system is the one constant, the "old reliable." It doesn't break and it is always available. You pick up a phone, and you hear dial tone. It just works. So, with that in mind, in the 1980s, if you were going to bring voice and data together, the PBX, with its high reliability, was a natural starting point.
Figure 1-1 offers an accurate view of voice and data integration as it was implemented in 1986. For those users who chose this solution, a single drop of wiring to the desktop was sufficient to handle both voice and data sessions. Many manufacturers offered the capability to connect voice and data desktop devices to the PBX, and of course, Integrated Services Digital Network (ISDN) Basic Rate Interface (BRI)—the basic rate interface with two information channels and a signaling channel (2B+D)—gave the industry an attempt at a standards-based way of delivering voice and data services to the desktop.
)
Figure
1-1 PBX Voice/Data Integration in the Mid-1980s
In Figure 1-1, the PC attaches to the telephone by means of a data terminal interface. PBX manufacturers had different names for this device. It was often called a datacom module, or a voice-data integration module, among other things. Regardless of the terminology used, this unit had a single function: convert the asynchronous stream of data into a format suitable for transport within either a single or dual timeslot. This device was found on both the upstream and downstream links; that is, at the desktop and at the host or computer location. In this manner, data devices were connected to the PBX and used the PBX as a means of connecting to a host computer, and shared the same wire as the phone connected to the PBX, resulting in cost savings.
So, on the surface, it looks like there was a solution almost 20 years ago for the "voice and data" industry. The solution worked as advertised, in terms of functionality and ease of use. It certainly introduced new desktop devices to the industry—such as the Cypress voice/data workstation and Cedar voice/data PC offerings from ROLM—and in many cases, was a cost-effective alternative to hard-wired data devices. However, this approach did have some drawbacks:
It was contention-based.
It lacked industry standards.
PBX architecture provided insufficient connection rates.
Contention
The concept of pooling, or contention, was a key component of the PBX-based voice and data strategy. Contention-based connectivity was both a benefit and a detriment to the convergence strategy in the 1980s. A contention-based solution allowed companies to deploy fewer ports to the host computers than users. In other words, there could be potentially hundreds or thousands of users contending for a limited number of ports. If a port was not available, then users were not granted access to the host computer. This was often the case with PCs or asynchronous terminals running some type of 3270 emulation package for access to an IBM (or compatible) mainframe. It was not uncommon to see a protocol converter emulate an IBM 3270 cluster.
The protocol converter, as shown in Figure 1-2, while hard-wired to the host computer, allowed PBX-connected devices to "pool" or contend for incoming ports. When all ports were filled, the users either automatically rolled to ports associated with the next protocol converter defined to the PBX (if available) or received a busy tone.
)
Figure
1-2 A PBX Allowing Data Workstations to Contend for Limited Ports on a
Protocol Converter
In Figure 1-2, four asynchronous workstations (VT100, PCs in async mode) contend for two slots on a protocol converter. The protocol converter converts the asynchronous data stream into a suitable format, such as 3270 for IBM System 370 machines, for presentation to the host computer.
This approach had both benefits and drawbacks. The main benefit was that companies were able to deploy lower cost asynchronous terminals (typically VT-100 type) instead of the more expensive 3278/3279/3179 devices. For users with personal computers, using less expensive asynchronous emulation cards instead of expensive 3270 emulator cards helped lower the costs to the organization. Also, because contention did not provide dedicated ports for every user, fewer "cluster controllers" were needed (protocol converters in this case) for direct access to the host environment.
The drawbacks, however, outweighed the benefits for many organizations. Because the goal was cost savings, as previously noted, each user did not have a dedicated port. For those users who only occasionally needed access to the host computers, this was a fairly decent solution. Yet, for those users who were accustomed to having access whenever they needed it, getting a busy signal was totally unacceptable. Here was a case where the traditional telephony way of handling a scenario (giving a user a busy signal) was, for some data users, out of the question.
Lack of Industry Standards
Another problem data users encountered with the PBX was a lack of standards. In the data world, it was necessary to adhere to certain standards. When connecting to a host, the Information Systems (IS) staff had to decide what kind of terminal to emulate, or imitate. So it was common knowledge among IS and telecom people that they might have to emulate a 3270 environment, a 5250 environment, a VT-100 environment, or an HP or Data General or Wang environment, and there were packages that enabled each and any of these emulations.
Utilizing the PBX, however, consideration had to be given to the type of port connectivity for desktop and host devices. Because of the lack of standards, the devices manufactured by one company weren't necessarily the same as the devices manufactured by other companies. So the data terminal interfaces that each vendor used were different, and each data manufacturer had to test against each PBX manufacturer without the benefits of standards.
Insufficient Connection Rate
However, more than anything else, the real issue companies faced trying to satisfy their data users when integrating into the PBX was the connection rate (line speed). Users who previously were accustomed to host-connected, or channel speeds (often in the 1–2 Mbps range), were now throttled down between 64–128 kpbs, which was the maximum connection rate that a PBX allowed. The reason for this was that a PBX allocated bandwidth in the form of timeslots, and each timeslot was, by definition, 64 kbps. This was the standard connection for voice. By providing two timeslots, data users were allowed double that connectivity.
For the "casual user" (a term created by the industry), this was generally acceptable. However, many users resisted the term. "There's nothing casual about my work requirements," they reasoned, insisting that their connectivity, although not continuous, was just as important and urgent. In the end, the slower speeds (which meant users watching their screens get "painted" line by line) and the busy signals doomed this approach. Contention and low connect speeds doomed voice-data integration in the 1980s. IP telephony eliminates these obstacles to convergence. Certainly, if IPT is going to work, it has to address the issues that grounded the movement to a halt in the early 1980s.