Despite availability of ISDN digital services and other latter-day enhancements, the Centrex market eroded steadily through the 1990s. According to one report, PBX's share of new installed lines grew from around 65% in 1990 to almost 80% (about 3,800,000 lines) in 2000 (Source: RHK, 2001, Courtesy Broadsoft).

During the same period, Centrex sales efforts by some major telcos (e.g., NYNEX) - previously aimed at government, universities, large institutions, and big business clients - became commingled (and, in some areas, de-prioritized in favor of) more general efforts to vend more minimally CLASS-enabled analog and ISDN lines (but not full-fledged, hosted PBX services) to SMBs. The percentage figures quoted previously may therefore represent not only "true Centrex lines" (i.e., lines used to provide hosted PBX services), but some CLASS-enabled POTS and ISDN lines, as well.

Why did premise-based telecom equipment run away with the market? Lots of reasons. Buyers wanted to share trunks among many employees, minimizing recurring charges. They wanted rapid, error-free turnaround on MACs. They wanted close control of system configuration. They wanted features, cute stations, enhanced voice messaging, and other adjuncts. They wanted to build call centers. Though Centrex solutions evolved in response, telcos implemented slowly, adopted non-competitive pricing, sold and supported poorly. At the same time, their attention was being drawn away from Centrex by T1 and PRI - more basic (and more profitable) services that better leveraged the copper plant, forced trunking buyers to think in terms of 24-line blocks, and answered growing demand for high-speed data.

Centrex lost mostly because telcos followed the money in other directions. But in another sense, Centrex lost because telcos were slow to grasp what PBXs were all about - what they were becoming. To them, a PBX looked like a channel bank - or a sort of "retarded, small Class 5 switch" - vaguely redundant; at best, an extension of the network. So, even as the PBX - like the PC - was evolving into "a platform for strategic enterprise communications applications," telcos were still treating Centrex like an extension of CLASS services - literally true, but psychologically way off the mark.

Today, this is finally changing. Telcos have come chastened through a long era of data-side buildout, a few wrongheaded flirtations with the idea of "the network as platform," and the increasing commoditization of raw bandwidth; and they've realized that the next wave of profits can only come from applications. Experience with the Internet has taught them new models for application hosting - models that accommodate significant access and control by end users, wrapped in attentive, proactive support. Exactly what Centrex needs to work.

At the same time, IP telephony technology has evolved with shocking speed, throwing a completely new spin on PBX vs. Centrex economics and operations. IP call switching and feature support are handled by a collaborating pool of self-configuring smart endpoint devices, proxies, database servers, rules engines, application and media servers, gateways. Except at the PSTN gateway, conventional, port-count, and switch-matrix-capacity-based notions of scale are gone - this new stuff, you scale by redeploying on a faster server, or adding a pizza box. Resources are placeless: Application components reside where it's cheapest and/or most sensible to put them - they can be anywhere with respect to the IP hard- and softphones used to access them and the client software or web browsers used to manage them (subject, obviously, to network conditions and bandwidth availability). DNS, DHCP, and other standard facilities let endpoints seek out servers, register, self-configure - even (increasingly) across firewalls.

Given all of the above, why have an IP PBX in your office? The major reason, at present, is because the gateways need to be there - attaching to conventional trunks (T1 or analog) at the demarc, and to whatever residual telecom wiring you retain to support legacy digital stations, analog telephones, and fax machines. But if you think about it, you might just as easily locate the trunk gateway elsewhere - say at a CO. (The "legacy phone interconnection" gateway, we assume, will become less strategic over time as legacy stations and analog phones start disappearing.)

In fact, this makes terrific sense. Who needs a trunk gateway in the office? It's a big chunk of vulnerable, mission-critical hardware - requiring climate-control and power protection, expensive to make redundant. In its premise version, it has a small port-count, so each port is expensive. Much smarter to let the telco keep it - own it - babysit it, leverage those economies of scale and co-locate the equipment with the PSTN facilities it connects to. Just give you a nice broadband pipe.

The server and proxy software? That can run either place. Makes more sense to run it on a server at the CO, though. It can be better secured there. It can run on the same subnetwork as related applications, so integration can be sturdier. The telco can afford more expensive, hot-swap, six-nines CPU hardware than you can, so it can be reliable. And (again relying on pure economies of scale) they can squeeze a lot more value into the package than you or the manufacturer could, acting alone. Things like: parallel FTP servers for fast endpoint-client updates. Things like: extensive and creatively updated online tutorials and "user community" facilities. Things like: tech support that knows your name.

In general terms, that's the vision. You own IP phones and wireless devices and basic data-network hardware. You rent everything else. And it doesn't matter: you still have real control. The telco, meanwhile, is your best friend. Because even though you're spending relatively little for a plethora of telephony features and related applications (UM, etc.), you're also buying their bandwidth, renting peak-time gateway capacity, getting your Exchange server hosted, your Internet access, your secure VPN, etc. All from the same source.

The devil, of course, is in the details. How carriers choose to implement IP Centrex - how friendly they feel they need to be to their own legacy installed base, how well they identify and promote the peak attractors, and how well they support real-world clients will all help determine whether this renaissance of Centrex can go the long haul.

Centrex IP vs. IP Centrex

Moving PBX-like functions to the network cloud was originally termed a "hosted PBX" but is becoming better known under two different names: Centrex IP and IP Centrex. Though some still use these terms interchangeably, significant players now acknowledge that they refer to different things.

Centrex IP is "IP-enabled Centrex." In Nortel's case, by adding IP line cards to Class 5 DMS switching systems, Centrex providers can support IP phones as part of conventional Centrex service offerings. Other telecom manufacturers add a separate, specialized gateway to the Class 5 switch, such as Lucent's iMerge gateway (see roundup), to drive the remote IP extensions and softphones. Lucent also adds a powerful softswitch to coordinate additional application support.