Compared with the average button-down computer-equipment show, the Computer Telephony (CT) Expo is wild. As evidenced by the inflaTable flying pink pigs out front, the expo’s tone is less serious than even an Internet convention.
But the CT business is dead serious in its quest to offer products and services that feature the best of both worlds from which it was born: the phone and computer industries. This hybrid business is growing rapidly. Computer Telephony Magazine and other industry sources estimate that it’s a $6 billion industry. Certainly, Lucent Technologies believes CT's a serious business; the company just spent $1.5 billion to buy Octel, a longtime provider of voicemail and CT equipment.
The CT industry is more mature and more standardized than it was a scant year ago. Although a gap still exists between computers and telephones, you no longer need original research just to make, for example, a Screen pop appear on a call center PC. (For more information about what the industry was like in 1996, see Alex Pournelle, "Computers and Telephones," Windows NT Magazine, June 1996.) To understand how this gap is closing so quickly, you need to look at CT’s parent industries.
A Peek into the Past
The telephone industry’s evolution closely parallels that of the computer industry. First came the mainframe era, in which AT&T provided all phone-related services, followed by the minicomputer era, in which the switch, or PBX, was key. In the minicomputer era, switch programmability was difficult at best and systems were closed to outsiders. As a result, if you wanted to bridge a call among three circuits (e.g., a conference call) or needed a queue for an operator pool (e.g., Automatic Call Distribution—ACD), you needed to install special switches or expensive add-on boxes.
Then came early computer telephone integration. Solutions were proprietary and difficult to implement, but they worked. Phone calls came right to a computer, which routed them to a phone or another switch. Meanwhile, switches and computers learned to communicate more effectively with each other, and PBX manufacturers opened their systems to outside control (as long as you could program in C).
Today, the CT industry closely resembles the PC industry. Traditional product categories have blurred, with hardware and software coming from many sources. The net result of this banquet of choices is confusion, especially for outsiders. At least when you bought a phone switch in the past, you knew which vendors offered them. Now, numerous vendor offerings cross the previously definite boundaries of "traditional" telephony products and services.
Sales channels have become confusing, too. At one time, telephony consultants were at arm’s length from both software and hardware developers; today, those same consultants might roll up their sleeves to punch down the cable, install the system, and maintain it. Distribution channels have become similarly confused. You can now buy phone systems through traditional computer sources and vice versa.
Despite the confusion, the CT industry has thrived, producing many innovations. These innovations fall into three categories: NT enhancements, applications and development tools, and hardware.
Windows NT: Now a Telephony Platform
Early CT systems ran on IBM’s OS/2, Sun Microsystems' Solaris, Novell’s NetWare, or MS-DOS. Today, CT systems are hosted on Windows NT with Telephony API (TAPI), NetWare with Telephony Services API (TSAPI), Solaris with JTAPI, or proprietary interfaces. The OS acts as an operator that directs and manages calls, sending commands through the API to the hardware. In such a setup, a Multi-Vendor Integration Protocol (MVIP), a Signal Computing System Architecture (SCSA), or another board-interconnect standard provides a mechanism for interconnecting many CT resource ports inside a PC-based platform. This interconnecting mechanism is one factor that differentiates CT solutions in a small office/home office (SOHO) from those in an enterprise.
Another difference between SOHO and enterprise CT is the level of control. Because most SOHOs have only one phone, first-party call control (i.e., the ability to control calls that come to your phone) is adequate. With first-party call control, you can dial a number, transfer a call, or put a call on hold. But because companies usually have numerous phones, they need third-party call control (i.e., the ability to control a call not directly connected to your computer). With third-party call control, a company’s computer can control an entire phone system, making applications such as automated banking services possible.
NT 4.0 with TAPI 2.0 made third-party call control a reality. But by industry standards, TAPI 2.0 is incomplete; enthusiasts have been waiting for the release of TAPI 2.1, which finally hit public beta in May. TAPI 2.1 offers client/server features for telephony control, a tightened standard, and important new features, such as Unicode support for internationalization, Active Control support, and a much better TAPI client manager. With TAPI 2.1, application-development tools and switch interfaces are smoother. (For more information about TAPI and other standards in the CT industry, see the sidebar "The Role of Standards in CT’s Growth".)
But the effectiveness of NT-based CT doesn’t depend only on how good TAPI is. This fact was evident at Microsoft’s booth at the CT Expo. Microsoft ran demonstration applications on a Northern Telecom Meridian 1 phone system. The demonstration featured Screen pops containing information about callers, ways to transfer calls, and techniques to control the phone system. And Microsoft’s demonstration barely scratched the surface of possible NT-based CT applications. (For more information on these applications, see Chris Bajorek and Alex Pournelle, "The Marriage of Computers and Telephones," Windows NT Magazine, September 1997.)
One nice benefit of attending an expo is seeing cutting-edge products. VenturCom’s Component Integrator (CI) definitely falls into that category. CI lets you embed NT or Windows CE on a no-hard-disk PC with realtime extensions. As a development tool for integrating, configuring, and building dedicated NT target systems, CI extends NT into places previously held by only UNIX and MS-DOS and offers obvious appeal to the CT market.
After the CT Expo, Microsoft introduced a cutting-edge product of its own: Routing and Remote Access Service (RRAS) for Windows NT Server. RRAS, formerly code-named Steelhead, is a networking software package that improves NT-based CT. (For more information about RRAS, see Mark Minasi, "Steelhead Swims into the Mainstream," Windows NT Magazine, August 1997.) RRAS connects two or more locations over Internet links or dedicated links (e.g., WAN). Because RRAS supports compressed and encrypted links, more MIS managers are likely to buy into the idea of using the Internet to haul their strategic packets. (For more information about Microsoft’s link-state routing protocol of TCI/IP networks, see Tao Zhou, "Steelhead’s OSPF Routing," Windows NT Magazine, August 1997.)
CT Applications and Development Tools: The Fun Begins
The CT field used to be all development tools and no off-the-shelf software. Now, the field has expanded greatly, so you can pick and choose from many products, ranging from complete SOHO software solutions to PBX front ends to development systems.
Many off-the-shelf applications are now powerful enough to remove the need for programming. For example, Active Voice’s PhoneMax is a SOHO solution with call control. Q.Sys’s CallProducer is an enterprise solution that displays a company’s call information at a master station. It interfaces a company’s existing phone system with its LAN and puts the front end on users’ PCs. Applied Voice Technology’s CallXpressNT is a unified messaging system for NT. It includes emails, calls, and faxes in one window. CallXpressNT will also read or fax email to you when you’re on the road. You can even remotely fetch faxes and forward them with voice annotation. (For more products that can perform unified messaging and other applications, see the "Computer Telephony Buyer’s Guide," http://www.winntmag.com.)
Naturally, complex solutions require a software development kit (SDK) and an advanced programmer to write the program. Many SDKs are Visual Basic (VB)-centric. You can write the program using the SDK libraries, which hide the complexity of the various functions (e.g., call control and voice record/playback). For example, Parity Software’s CallSuite uses VB, but also supports Visual C++ and Delphi (which is appealing for enterprise-level applications). Parity Software’s TAPI Starter Bundle is a cheap way to get your feet wet in CT and comes with a 700-page PC Telephony book.
Artisoft, famous for LANtastic, now sells Visual Voice, one of several VB-centric application creators available. Visual Voice has a text-to-speech option for playing back those "Your account balance is $17.31" messages from your bank. In addition, Visual Voice tracks call accounting (Who’s calling whom?) and call tracking (Who did this guy call in our office?) information, which can often help sell the application to the boss. A pro version for multiline call-processing boards is also available.
Some application creators use languages other than VB. For example, Expert System’s Ease for Windows NT is an integrated development environment (IDE) that uses its own low-level language and hooks to C/C++ for complete customization. With the IDE, you complete scenarios and provide triggers, actions, and data in a standard Windows framework. Ease for Windows NT has a runtime environment that sits on NT once you complete the installation.
Sometimes products cross category lines, such as Octel Communications' Unified Messenger. This product lets you retrieve voice mail messages from any phone or computer and send voice email attachments. It’s both a traditional computer program and a CT application. The idea of unified messaging and a universal inbox is now becoming reality.
Like the PC industry, the CT industry is fractured. Some vendors provide call-processing cards, while others offer PBXs-on-a-card. Meanwhile, some companies sell complete systems. All solutions work, to varying degrees, with either their application development tools or third-party tools. Industry leaders exist, but none dominates.
Although the structure of the CT and PC industries is similar, big differences exist between traditional computing and CT applications. One significant difference is the servers. CT applications run on big hardware. The show was full of CT servers right out of the industrial-computer world, complete with rack mounting, big fans, hot-swap disks, 10 to 20 slots, passive backplanes, CPU and RAM on daughter cards, redundant power supplies, and temperature alarms. Surprisingly, instead of seeing IBM, Compaq, AST, and other well-known PC names at the expo, I saw names such as Dolch, InterLogic, Ziatech, and Recortec. None of the big PC names makes enterprise-size CT servers, except Digital Equipment, which offers Alpha servers for that purpose.
Whoever the maker, the CT server usually has call-processing cards and full-length ISA boards stuffed with analog components and digital signal processor (DSP) chips. Low-density cards, such as the 4-line Rhetorex RDSP/9432, have individual phone-line connections out the back of the board and some amount of onboard CPU. High-density boards, such as Dialogic’s D/240PCI-T1, might take an entire T-1 line worth of calls. No matter the cards’ density, all listen for ringing, play a greeting, listen for touch-tone or rotary digits (or the caller’s voice with the appropriate software), and then route the call. Some cards might then play another digitized message, offer text-to-speech conversion, or bridge conference calls.
In the world of CT servers, one rule holds true: the smarter the board, the less burden on the host CPU. The D/240PCI-T1, for instance, has three 66MHz Motorola DSP56002 DSPs and two Intel 486 GX CPUs for call processing. Obviously, for more smarts, you can expect to pay more money. (If you’re not using an SDK such as Visual Voice and CallSuite, you need to make sure that your application-development software knows how to harness these smarts.)
The call-processing boards that I’m mentioning are but a tiny sample of the available boards. Rhetorex, for instance, makes more than 20 boards. Dialogic offers more than 30, ranging from 1 line to 120 lines for ISA, PCI, and VMEbus. Some of these boards plug directly into a PBX. In addition, many Dialogic workalikes exist, such as NewVoice and Bicom. You must ensure the boards you select will support the software you want to install.
But you don’t need to build your system from components. If you want a total solution, try an UnPBX, an NT-based phone system with automated attendant, voice messaging, ACD, and email functions. One exciting UnPBX approach, for example, is Sphere Communications’ Sphericall software and hardware product. Sphericall puts a switch in an NT Server. But it can also use asynchronous transfer mode (ATM) networking to combine LAN and voice traffic on the same wires. You just need Sphere Communications’ PhoneNICs to connect each phone to the host PC. The company also offers PhoneHubs to add plain old telephone service (POTS) phones without ATM. Most of the call handling is on Screen, so you don’t need expensive phones. (For more information about Sphericall, see Chris Bajorek, "What’s Happening in Computer Telephony," Windows NT Magazine, January 1997.) Sphericall blurs the line between computers and telephony. In fact, Sphericall received more attention at NetWorld+InterOp than at CT Expo.
Despite all the board-level, UnPBX, and hybrid solutions, don’t expect the traditional switch to dry up. Switches have redundancy and reliability that make PCs look like toy computers. (You can’t even put a card in a PC without turning it off.) In addition, many companies already have large investments in switch equipment and telephones. So, instead of companies replacing their switch equipment, they will most likely use computers to more easily control the switches.
Predictions Through a Foggy Crystal Ball
Two events will affect CT in the near future. One event is the release of NT 5.0. (For more information about the new NT 5.0 capabilities that will affect CT, see the sidebar, "Windows NT 5.0 and CT.") This new version of NT will make CT applications, especially Internet-based ones, easier to develop—and that brings us to the second event: the growing reliance on the Internet. The widespread popularity of the Internet will greatly affect both CT and NT. Companies will need to use both to help them meet their customers’ rapidly changing needs. For example, companies will need to integrate Internet phone-like plugins into their customer service and tech support centers. They will also require page-push features to show customers the right Web page. Because more voice- and videoconferences will take place over the Internet, companies will need to install full-time (and large) connections. Companies will also need telephone systems with an Ethernet jack so that they can make and receive voice-over-IP calls directly. Finally, Internet delays will make extranets to vital customers and suppliers more important.
In the more distant future, perhaps NT 6.0 will have more smarts for CT. But the independent CT application-development tools and CT hardware market aren’t going to fade away. Like dreams, both will change in unpredicTable and exciting ways. Connecting and using all these new innovations will get easier for SOHOs and FORTUNE 500 organizations alike. I can’t wait for next year’s CT Expo show to see how.
The Role of Standards in CT’s Growth
Standards have been incredibly important to the CT industry’s growth and are shaping it today. Interoperability is one area in which standards play a critical role. Companies want to buy software and hardware components from multiple vendors, install them on a Windows NT server, and have them all work together in call-automating harmony. For example, you can buy a voice card, fax card, voicemail software, and fax server package from different vendors, and through the magic of standards, they peacefully coexist. They can even work as an integrated messaging solution with Microsoft Exchange Server and Outlook.
Vendors now have many standards to choose from. Open Database Connectivity (ODBC), SQL, Telephony API (TAPI), Telephony Services API (TSAPI), Messaging API (MAPI), H.323, S.x00, and Voice Profile for Internet Mail (VPIM) are but a few. (For more information on these standards, see Chris Bajorek and Alex Pournelle, "Computer Telephony Terms and Technologies," Windows NT Magazine, September 1997.)
In the last year, NT became a pivotal standard for CT, providing a common software platform for CT manufacturers. In fact, Microsoft has been a driving force in CT by providing the following standards:
• PC-based OS (MS-DOS, Windows 3.x, Windows 95, and NT)
• ODBC for database access
• Messaging API (MAPI) for messaging
• COM/DCOM for distributed object-oriented development
• Application development tools (primarily Visual C++ and Visual Basic)
• TAPI and related API standards
While Microsoft’s CT focus and momentum waned a bit during its full frontal attack on Netscape, the company recently updated its CT efforts with COM-enabled TAPI 3.0. Microsoft has also been busy with an NT networking standard important to CT: RRAS (formerly code-named Steelhead) for NT Server.
When multiple corporate sites have their own CT systems, a natural need exists to network them. For example, if you want to send a voicemail message to a remote office, you need to send that message from your local system and have the CT system automatically transfer it to the appropriate remote mailbox. With RRAS, you can establish server-to-server connections that provide a conduit for the routing of email, voice, fax, data, and video with equal ease and efficiency. Of course, you need application code to implement the routing function, but with NT making the network connections, CT application developers have one less piece that they must create themselves. The same update also makes it easier to create IP-to-PSTN gateways because establishing the server-to-server connection is an intrinsic NT capability.
Windows NT 5.0 and CT
Core features in Windows NT 5.0 will spawn new computer telephony (CT) products and capabilities. Although NT 4.0 is already a highly capable CT platform, NT 5.0 will make CT applications, especially Internet-based ones, easier to develop. Here are several new NT 5.0 capabilities that will affect CT.
Active Directory Service: The most strategic addition to NT 5.0 will be the built-in Active Directory Service. It will enable many user-addressing features for communications routing, all of which NT will handle. One such feature will let you publish controlled user information in a global directory entry. The global entry will let anyone on the Internet find you.
Active Movie: New data-streaming protocols will simplify implementing Internet telephony applications. These protocols will, for example, make IP-to-Public Switched Telephone Network (PSTN) gateways easier to implement and appropriate for all realtime communications streams, including voice, fax, and video.
COM/DCOM support: NT 5.0 will offer support for Component Object Model/Distributed Component Object Model (COM/DCOM). With this support, NT can provide scalable object-oriented solutions.
Hot-swap: NT 5.0’s hot-swap feature will let users exchange NT Server components without shutting down the server. This new feature will most likely prompt PC vendors to begin supporting hot-swap capabilities, although they won’t address hot-swapping PCI cards.
Nonstop computing: On the software front, when Microsoft releases the Wolfpack clustering program, more software and hardware companies will use nonstop computing for failover in telephony. Thus, NT 5.0 users will have nonstop computing capability.
Plug and Play (PnP) support: On the hardware front, NT 5.0’s PnP architecture will slowly make CT hardware easier to install compared with installation on those systems in which the call-processing boards have ISA architecture. But the switch to PCI will be slow until CT becomes more do-it-yourself. In addition, PCI’s three to five cards-per-PCI-bus limit will keep ISA from fading away for another year, at least for applications requiring 8 to 10 boards. This situation will put CT in a separate box and out of the file server, especially in large installations (120 lines or more).
ReSerVation Protocol (RSVP): Microsoft will implement this quality-of-service protocol for the Internet as part of NT 5.0. RSVP will enable realtime communications that will be far less susceptible to dropouts.
Telephony API (TAPI) 3.0: This version of TAPI will merge call control with media streaming and control. TAPI 3.0 will have features that make Internet telephony applications easier to implement. For example, a COM layer will be put around TAPI 3.0, so developers can use their language of choice (e.g., Java, C++, or Visual Basic).
Virtual private networks: NT 5.0 will provide efficient server-to-server routing over virtual private networks, providing low-cost CT bandwidth. No router hardware is necessary. These connections will provide realtime communications for all media types. (This capability is already available for NT 4.0 through RRAS, which Microsoft released this summer.)
Contact: Q.Sys • 513-745-8070|
|CallSuite, TAPI Starter Bundle|
Contact: Parity Software • 415-332-5656|
|Contact: VenturCom • 617-661-1230 or 800-334-8649|
Contact: Dialogic • 973-993-3000 or 800-755-4444|
|Contact: Active Voice • 206-441-4700|
|Contact: Rhetorex • 408-370-0881, Ext. 1|
|Routing and Remote Access Service for Windows NT Server|
|Contact: Microsoft • 206-882-8080|
|Contact: Octel Communications • 408-324-2000|
|Contact: Artisoft • 617-354 0600 or 800-914 9985|