January 1998

BY DAVE CORLEY

A revolution is overtaking the data communications market. Voice and video services, so long presumed to be the sole domain of circuitswitched and cell-switched networks, are banging loudly at the door of packet-switched networks. The data communications industry is about to answer the door in a big way. A wide range of new products and services is systematically overcoming many barriers that have prevented acceptance of isochronous service delivery on packet networks. Several vendors’ products are beginning to make the converged LAN PBX concept a reality. Voice over IP gateways are available from vendors such as NetSpeak, VocalTec, and others. Ethernet PBX phones and PC hardware and software product combinations from PhoNet, QuickNet Technologies and Microsoft extend the data network to the user’s phone device. This article outlines the remaining barriers to total LAN PBX integration, and why they are disappearing.

THE “—ABILITIES”
Enterprise communications managers evaluate voice service systems based on their “—abilities.” Value is placed loosely among several categories of “— abilities,” including affordability, manageability, scalability, redundancy/faulttolerance, reliability, functionality, interoperability, and usability. Every user, administrator, purchaser, and manager of these systems will assign their own ideas of the value of these abilities based on their own motivations and desires. The manager charged with making the decision to purchase and deploy a system must weigh each person’s value assessment, cast aside technology religions, and move forward with deployment decisions. Within the next two to five years, one of those decisions will likely be whether or not to widely deploy a stand-alone LAN-based PBX as an adjunct or outright replacement to the traditional circuit-switched PBX. Let’s look at that decision in light of the abilities mentioned above and the evolving technologies that are making the LAN PBX possible.

PBX HISTORY
With telecommunications deregulation came a flood of Private Branch Exchanges. These systems were initially targeted at large enterprises that were paying exorbitant rates for mediocre voice service from monopolistic public telephone organizations. The PBX provided communications managers the opportunity to improve service, manage their own technology destiny, and reduce recurring telecommunications costs. A simplified view of the architecture of the PBX is presented in Figure 1. A proprietary operating system and proprietary applications run on an expensive processor. A circuit-switch is controlled through a master/slave relationship with its master processor. Station devices in the form of analog or proprietary digital phones are connected to the circuit-switch through proprietary hardware line cards. The public telephone network is connected to the circuitswitch through proprietary hardware trunk cards. If a competing PBX manufacturer developed a knock ’em dead hardware component which enhanced its vendor’s PBX capabilities by 1,000%, you can be sure that component could not be added to another vendor’s PBX. Hardware component interoperability among PBX manufacturers was non-existent. Similarly, internal software components are not interoperable. For example, the operating system on a Lucent Definity does not operate on a Nortel Meridian PBX. Within the past ten years, this picture has changed to some extent. Vendors such as Dialogic, Natural MicroSystems, and Rhetorex market hardware components that offer the same functions as PBX switches, line cards, and trunk cards. Operating systems such as Microsoft’s Windows NT and Sun’s SCO UNIX provide interoperable software application interfaces on which telephony applications have flourished.

A new generation of PBX manufacturers has developed scaled-down versions of the PBX, loosely termed the “unPBX,” by combining communications hardware and software components into PCs. These PC PBXs are off-theshelf units with published application programming interfaces — developers only need the skills to write the applications that enable the services provided by the combined components. They don’t have to know how to design hardware or low-level drivers. The result is products delivered to the market more quickly. The developer can rapidly prototype a concept PBX, then alter its behavior quickly according to the customer’s needs. However, the underlying architecture of the PC PBX does not differ from that of the mainframe PBX. It’s in a less expensive and smaller box — in most cases a PC. The master/slave relationship between processor and switch and line/trunk cards persists. PC PBX manufacturers also struggle to provide a scalable system beyond a few dozen users. Once the slot capacity of a PC is consumed, most PC PBX users cannot extend PBX features across multiple boxes. So a user on Box A may not be able to transfer a call to a user on Box B. Finally, it’s not tightly integrated with the data network. Phones, trunks, and switches are circuitswitched. Data network components are not. Computer-telephony integration (CTI) links between data services computers and PBXs have been used over the last decade to bridge this gap. Lack of focused standards has slowed acceptance of the CTI approach.

THE NEW BREED
A new breed of PBX, the LAN PBX, brings all components into the packet network. There are three essential elements to the LAN PBX: terminal devices, call management software, and gateways (Figure 2). In this model, the terminal device is an IP node with a TCP/IP stack and an IP address. There are two general classes of terminal devices. The first is a PC with a software communications application such as VocalTec’s Internet Phone or Microsoft’s NetMeeting. Sound cards or dedicated, special purpose PC cards such as QuickNet Technology’s Internet PhoneJack, perform audio encode/decode and acoustic echo cancellation. The second terminal device is an Ethernet PBX phone like the one from Selsius Systems — an intelligent PBX phone that connects directly to a 10Base-T network.

The second element of a LAN PBX is call management software located in a call-processing server. Interoperability demands that this software be compliant with International Telecommunication Union-Telecommunication Sector’s (ITU-T) H.323 body of standards. Scalability and redundancy would suggest that the servers are capable of wide distribution on the IP network and capable of managing shared resources. Within this software component resides the intelligence to enable supplementary PBX services such as call forward, call transfer, and multiple line appearances. The third and final element allows LAN PBX users to communicate with terminal devices connected to cir-cuitswitched networks. These devices, called IP to circuit-switched gateways, bridge the packet network to the PSTN and the PBX. Other gateway devices allow direct connection of legacy analog FAX machines, POTS phones, and modems directly to the LAN. Contrasting this architecture with that of the PBX and PC-PBX, the hardware circuit switch of the PBX/PC-PBX is replaced by existing data network components – hubs, switches, and routers. Since terminal, call processing, and gateway devices are TCP/IP devices, the network transport becomes irrelevant. The network cloud can be composed of cell-, packet-, or circuitswitched distribution components. The line cards of the PBX and PC-PBX do not exist in a LAN PBX. PCs and Ethernet phones connect directly to the data network and do not require special interface cards. Finally, the trunk cards of the PBX/PC-PBX are replaced by the IP to circuit-switched gateway. Elimination of the separate switching infrastructure, line cards, and the use of TCP/IP protocols deliver advantages with respect to manageability, affordability, and scalability that are discussed below. The LAN PBX enterprise telecommunications solution speaks to the essential “—abilities.”

USABILITY
A phone is familiar. We’ve all used phones to make a date or to make a deal. It is a natural candidate for a voice terminal. It’s a well-understood device in this high-tech world. It’s also portable. An Ethernet phone with its own automatically assigned IP address may be moved from jack to jack within an organization without any administration. The phone always retrieves its configuration when it’s plugged in to the new jack. For example, if a user expects an urgent call during an important meeting, he could grab his phone and take it with him to the conference room, plug it in, and wait for the call. Because of its intelligence and the fact that it’s connected directly to the IP network, an Ethernet phone can perform special operations such as initiating video or whiteboard calls from a companion PC with a single button-push on the phone. PC communications applications are finding more favor with users familiar with high tech applications. For example, NetMeeting, Internet Phone, and NetSpeak’s WebPhone provide easy-to-use interfaces that allow users to manage voice, data, and video communications. The host PC processor, an installed sound card, a special purpose telephony interface card, or a combination of these devices perform audio encode/decode and echo cancellation functions.

MANAGEABILITY
Reduced total cost of ownership (TCO) of business network systems is driving enterprise purchasing decisions. Network managers understand that recurring administrative costs of installation, training, operations, and maintenance are the most significant component of TCO for these systems. Tools are now available that automate tasks and reduce the cost of administering data networks. For example, BOOTP and DHCP protocols allow automated IP address assignment to voice terminal devices. Automated registration of these devices with their call processing servers allows terminal configuration information to be downloaded. TFTP protocols allow software upgrades of diskless voice terminal devices on command from a remote management console. The cost of moving or upgrading a terminal device is lowered as a result. Finally, the use of World Wide Web protocols and advanced Simple Network Management Protocols allow an administrator to configure and monitor the PBX devices and database remotely from an almost universally available application — the Web browser.

INTEROPERABILITY
Not every vendor gets it right. Most will get it right with some of their products and fail with other components of that same product line. The enterprise is normally a mix of vendor products that must interoperate, so the importance of interoperability standards and their implementation and compliance verification is clear. The ITU-T, through the H.323 set of protocols, and the International Multimedia Teleconferencing Consortium (IMTC), through its Voice over IP Forum, have forged sets of evolving standards that govern the behavior of interoperable components. In particular, the H.323 concept of a gatekeeper equates roughly with the call-processing server. The standard also defines gateway interoperability. Protocols within the standard govern signaling, device registration, and capabilities negotiation among interoperable components. No one vendor can build all of these components to perfection. An enterprise network administrator must choose among the best and manage in an interoperable environment. This requirement dictates that individual components be interoperable.

FUNCTIONALITY
User and administrator functions must not be compromised because of the move from a circuit-switched to packet-switched communications network. Business users count on PBX and key system functions such as dialtone, transfer (with and without consult), hold/release, forward (all, busy, no answer), mute, last number redial, callback, call park/pickup, speed dial, multiple line appearances on single terminals, multiple terminals with single line appearance, vacant number intercept, pacifier tones for local DTMF/key operations, and DTMF generation for legacy Interactive Voice Response system activation. Administrators must have functions to manage flexible and distributed dial plans, call detail records, station message detail records, remote database configuration, and automated address translation. Unified messaging, attendant, and conference bridging are adjunct application requirements shared among users and administrators.

AFFORDABILITY
The transition to a converged network LAN PBX must be cost effective. Both fixed and recurring costs must be consistent with or lower than circuitswitched PBXs. LAN PBXs offer this fixed cost affordability through savings in wiring, switch, and interface card expenses. Recurring costs are saved through reduced moves, adds, and changes. Use of common skills among MIS personnel leads to lowered training and personnel management costs. Traditional PBXs, by the nature of their proprietary architecture, require specific technical skills taught by their vendors. These skills are not easily transferable to other areas. The client/server and TCP/IP nature of LAN PBXs (as well as their deployment on relatively common PC platforms and operating systems) equate the skill sets required of LAN and telecommunications systems managers. “Bell-heads” and “packetheads” can learn from the same teachers. The pool of potential workers from which a telecommunications administrator might be hired expands to include most data network professionals.

SCALABILITY
A LAN PBX architecture can affordably solve communications problems for enterprises serving two to thousands of users. The ubiquity and flexibility of the TCP/IP suites allow distributed processing, shared resource management, and distributed resource management concepts to create virtual LAN PBXs spread across the globe. The software nature of the call processing components of the system and the low cost of microprocessors means that call processing can be spread across multiple subnets. Several hundred terminals can be logically assigned to their own callprocessing server. Inter-server signaling methods such as QSIG can be employed to extend feature transparency between terminals served by separate call processing domains. The benefit is a virtual PBX, where components require only TCP/IP connectivity to communicate. An example of a distributed PBX is presented in Figure 3. Call processing (CP) servers are distributed among central and remote offices in the example enterprise. Terminals at any System reliability is affected by component reliability. Vendors of terminals, hubs, switches, routers, PCs, applications, and operating systems are beginning to heed the call for more reliable components.

office can be assigned to any CP server. The CP server need not be within the enterprise. Gateways for interconnection to local PSTN or PBX devices provide a path for calling stations on the PBX or PSTN. These gateways are resources shared among the users. CP manages the resource sharing. These are also distributed among enterprise offices. Other shared resources are distributed among the office IP clouds. The common transport technology among these networks is TCP/IP. In theory, the LAN PBX is infinitely scalable, through CP or shared resource component addition.

REDUNDANCY AND RELIABILITY
TCP/IP and advances in database technologies such as automated replication provide a path to redundant CP service. Any terminal or gateway device can point to a primary, secondary, or tertiary CP server. If the terminal loses TCP/IP communications with its controlling CP server, it transfers control to its secondary CP server. Calls are preserved through the failure since the voice stream passes directly between terminals. Other network devices display similar redundant component behavior. System reliability is affected by component reliability. Vendors of terminals, hubs, switches, routers, PCs, applications, and operating systems are beginning to heed the call for more reliable components. Every part of it has to work, even if the lights go out in the building. Terminal components pose a particular problem with respect to power outages. It’s too costly to place battery back-up power supplies on every terminal. Instead, power to terminals, specifically Ethernet phones, must be delivered from a cost-effective, centralized power source that utilizes a UPS (Uninterruptible Power Supply). Products will be available in early 1998 to deliver remote power to low-powered terminal devices.

Get ready for the LAN PBX revolution. It is upon us now and will break down the door if we refuse to politely answer. It will be without some social graces at the beginning, but expect it to conform to your tastes, expectations and civilities as the new millenium dawns.

Dave Corley is director, product management of Selsius Systems. Selsius Systems designs, develops, and markets LAN PBX components and systems. The company is located in Dallas, Texas. Its products include the SelsiusPhone, a family of PBX phones that connects directly to an Ethernet network; Selsius-CallManager, a Windows NT-resident software PBX; and SelsiusAccess, a family of IP-to-circuitswitched gateways with analog and digital interfaces. For more information, contact the company at 800-8558200, or e-mail [email protected].