For years, Centrex has been a good, solid business for the incumbent LECs and Baby Bells, but it's never caused all that much stir industry-wide. For non-users of the service - the majority of medium-sized businesses out there - it may be surprising to learn that the market for Centrex services actually comprises around 15% of business phone lines overall. Still, this slice of the pie has remained relatively stable over time, and its modest growth has not posed any significant threat to PBX sales. Centrex has its niches: It can be found in government networks, on university campuses, among businesses too small to cost justify a PBX or key system, and in some very large enterprises which might also own their own phone systems. Outside these pockets, however, carriers have made little effort to market the service, and companies have shown little interest in buying it.
So what has gotten so many people so excited about the possibility of delivering Centrex over IP networks? On one level, the question answers itself. Practically any service with "IP" in its name, particularly if it's not yet been deemed a commodity, has the ability to generate buzz - sometimes justifiably, sometimes not. Add to this the fashionable label of "hosted services" - which, of course, could be applied to the Centrex model - and it isn't hard to see where the excitement is coming from. But beyond its external dressing, the actual value, or even the basic definition, of IP Centrex is rather more difficult to pinpoint.
One complication stems from the fact that one of the primary and most lucrative target markets for IP telephony in general - small- to medium-sized businesses - is precisely the category that has traditionally opted for premise-based phone systems, as opposed to Centrex services (view table) Unless IP Centrex can expand from today's existing Centrex user base, and begin to capture market share from PBXs, its potential to stimulate competition and drive increased revenues will be limited. Related to this is the issue of feature differentiation: Until now, Centrex services have consisted of a common set of standard features, with little variance among offerings from different providers. This is one reason why CLECs have not had more of an impact on the Centrex market. The ability to offer differentiated services, beyond what current Centrex or even PBXs provide, is a crucial yet still somewhat vague requirement for IP Centrex. And perhaps to an even greater degree than the features themselves, the acceptance of IP Centrex among a broader range of customers will depend on how creative providers can become in marketing, selling, and supporting the service.
CO Roots
The appeal of conventional Centrex, as well as its shortcomings, is by now relatively well understood. Small businesses like the fact that it requires little to no upfront investment in equipment or installation, and the fact that it can be leased on a month-to-month basis, which protects against lock-in for companies that are changing or growing rapidly. Centrex gives these customers the basic business features they need - DID numbers, four-digit dialing, call forward, call hold, etc. - without making them buy and manage their own phone system. Large businesses, on the other hand, who typically have longer Centrex contracts and may also have equipment deployed on premises at their main sites, like the fact that Centrex, as a central office service, can scale almost indefinitely. It can also be more cost effective for getting service to multiple locations than connecting PBXs with tie lines, or leasing off-premise extension (OPX) lines.
Where Centrex tends to fall short of the PBX alternative is in the features that it offers, as well as in the customer's ability to control the service. Part of the problem is that Centrex offers only two options for delivery - POTS or variants of ISDN BRI - and thus two options for the type of terminal used - cheap, dumb analog phones, or somewhat smarter, more expensive, proprietary or open-architecture ISDN phones. For the cost-sensitive customer, the former is the only choice, while those willing to shell out the extra money still don't get a lot of bang for their buck. The other problems, of course, are the difficulty of programming new services into a class 5 switch, and the user interface challenges of a standard telephone (star codes, flash hooks, etc.). Even more significant is the fact that a Centrex user relies almost entirely on the service provider for configuration and management (despite some advances in areas like call accounting). While having someone else manage the service is arguably a key value of Centrex, there are certain areas - moves, adds, and changes (MAC), for example - where it just makes sense to extend some control to the user. Because of the underlying switch and network architecture that supports today's Centrex services, however, this has been impossible to achieve in any efficient manner.
Perhaps the most striking fact about the Centrex market as it stands today is that, for all its apparent stability, there seems to be very little beyond price and the legacy of an organization that motivates customers to choose Centrex, and even less that ties them to a particular vendor. One study, for example, shows "lower cost" as the primary reason why customers elected a Centrex service over a PBX, followed closely by perceptions of greater reliability and "historical reasons" (see Figure 1). What's interesting about this comparison, however, is not that it reveals any value unique to Centrex offerings, but exactly the contrary: The reasons cited in favor of Centrex almost directly mirror those in favor of the PBX.
The ambivalence that this sort of data reflects, on both sides of the CPE/hosted-services fence, is in fact a double-edged sword. On one hand, it confirms the notion that there has been little differentiation in terms of Centrex features, and that carriers have not had much success in "branding" the service to gain customer loyalty. Note that both "feature functionality" and "vendor image" are relatively insignificant factors in the Centrex customer's decision, which seems to indicate that customers are choosing their Centrex providers largely by default. On the other hand, this same indifference appears on the PBX side, which, if accurate, means a significant opportunity for Centrex offerings that can differentiate themselves to capture market share.
The fact that more customers point to cost than any other factor seems in some sense to be a red herring. Yes, a Centrex offering must continue to be cost competitive. But to date, there simply has not been any other concrete basis for comparing among Centrex services, or between Centrex and other alternatives.
And why will IP succeed where more conventional, circuit-switched Centrex hasn't? Most vendors and service providers involved in IP Centrex like to stress the fact that an IP-based softswitch, as opposed to a traditional class 5, makes it easier to develop and deploy new applications. But just adding new features - examples such as unified messaging, find-me/follow-me, and, more recently, presence services are often discussed - isn't likely to convince historical PBX customers to switch to a different model. What we think will drive customers to consider adopting Centrex, or to consider changing Centrex providers, has more to do with the inherent capabilities of IP itself, and with the strengths of an outsourced model for service provisioning.
The advantages of a "one-wire" infrastructure, for example, in terms of both cost and management are a clear step up from current Centrex, which requires a separate physical circuit for each user's line, and makes a static association between a copper pair and a phone number. With IP, all voice connections are delivered over a common pipe to the customer premise, and distributed to IP addresses across a LAN/WAN. In this model, adding an additional user, or moving an existing one, becomes easy. For customers with multiple locations (69% of all Centrex subscribers according to last year's National Centrex Users Group survey), the same capabilities can extend to the wide area network, where common dialing plans and feature groups can be shared across a voice VPN. In conventional Centrex, it is only recently that multi-site enterprises have been able to get networked service in a single metropolitan area, while more remote sites must be served from different, non-integrated central offices.
These same characteristics apply as well to a premise-based, IP-PBX model as they do to IP Centrex, and there is no doubt that both will be prevalent in the future. But we think the argument that current PBX customers like to own their phone systems just for ownership's sake is a myth. If a carrier could offer these customers a viable way to outsource their telephony, without incurring additional cost or sacrificing functionality, the majority of them would jump at the opportunity. As application service providers (ASPs) become more visible in areas such as web hosting and enterprise software (CRM, ERP), we think telephony will fall in line as another mission-critical, hosted application.
Two Paths Diverge
IP Centrex has not yet really emerged as its own market, despite some early offerings from startup service providers. For the most part, the industry is projecting beta tests to happen during the first half of this year, with live deployments getting underway in quarters three and four. At the same time, vendors and service providers together are already pursuing at least a couple of different deployment models.
The clearest divide to emerge thus far is between greenfield providers and the incumbent or the more established competitive, LECs. The latter group, ILECs and CLECs, typically have class 5 switches already deployed in the markets they serve, and in many cases are offering Centrex services today. The priority for these providers is to transition to an IP-based architecture (not just for Centrex, but for call switching in general), while holding onto their existing subscribers and defending their market position against both direct competitors and new IP-PBXs. In addition, IP Centrex could represent a growth opportunity for these carriers, who are watching the margins being squeezed out of local dialtone (just as they were out of long distance) and who are eager to provide broadband enhanced services. The greenfields, on the other hand, need to match the offerings of the ILECs with a significantly lower capital investment, and move quickly to develop features and applications that will help them target vertical markets.
Not surprisingly, the vendors who will supply software and hardware platforms to these various groups are similarly divided. At one end of the spectrum are Lucent's and Nortel's IP Centrex solutions, which leverage core technology from those companies' CO switches, and take a phased approach toward transitioning to IP. At the other end is the ever-expanding collection of softswitch and application infrastructure players, for whom the services as well as the switch are built from the ground up in an all IP environment. These lines, of course, are blurry, and within each broad camp one could draw countless further distinctions, but they can at least help indicate some basic differences in philosophy, which in turn lead to different conceptions of what an IP Centrex service will look like.
Migrating To Packets
Incumbent carriers certainly haven't been blind to the changes occurring in their competitive environments, and the prospects of IP telephony have already had an impact on their Centrex strategies. Lucent and Nortel, who together own the overwhelming majority of the class 5 switch market, and by extension are the biggest suppliers of Centrex equipment, have both put forth migration strategies for their switches that specifically address IP Centrex as a requirement. Among the customers who have already agreed to trial IP Centrex services based on these vendors' new platforms are SBC, working with Lucent, and Bell South, working with Nortel.
Lucent (Murray Hill, NJ - 888-4-LUCENT) divides the IP evolution plan for its 5ESS switch into four phases: IP-based access to 5ESS features from the line-side of the network; trunk-side interfaces for CO-to-CO transport via IP; migration of the switch itself to a decoupled, software-based architecture; and, finally, opening up the programming interfaces to the switch for third-party application development. The first phase, as it pertains to Centrex, involves deploying the Centrex Feature Gateway (CFG), a product that will be part of the trial Lucent has planned with SBC in April. Here, the class 5 switch processes calls exactly as it would in a conventional Centrex environment, but hands them off to the CFG, which does protocol conversion between IP and ISDN. At the customer premise, IP phones or an analog phone gateway hook up to the LAN (today these speak to the switch gateway using the H.323 protocol), and traffic is routed out across the local loop over a high-speed data connection.
At this stage in the evolution, the only major feature differentiation comes from the IP phones themselves (for which Lucent will partner with various third parties), since the Centrex features resident in the switch are exactly the same as before. But, as indicated earlier, the cable management and MAC problems that a LAN telephony model can solve are not insignificant attractors to this new model. For this same reason, Michael Meyers, Lucent's director of product management for 7R/E Solutions, points out that transitioning their access lines to IP can help carriers defend their Centrex customers against IP-PBXs, and even begin to take business away from the PBX vendors.
"Certainly these carriers want to protect their current Centrex base, and IP PBXs pose a serious competitive threat," says Meyers. "But the fact is that Centrex still represents a relatively small percentage of the total number of business phone lines. So they're also looking at this as a way to expand their market. A lot of customers are getting ready to upgrade their PBXs right now, and this is a chance for the Centrex providers to gain market share."
From a high-level perspective, Nortel Networks(Brampton, ON, Canada - 905-863-0000) is taking a similar approach with a program it calls Centrex IP. Today, Nortel makes a gateway card that sits inside the DMS-500 switch, doing protocol conversion. An external server runs gatekeeper software to handle addressing and authentication of client devices. One difference from Lucent is that Nortel is making its own IP phone sets, currently available in one model, the i2004, but soon to expand into others. Nortel also has a unique implementation of H.323, which uses proxy servers to mediate between the phones and the gateway, rather than fully embedding the protocol stack in the end device. The advantage, the company says, is that these proxy servers can be easily upgraded to support other protocols, such as H.248/Megaco or SIP, without having to swap out the phones themselves.
Nortel's larger carrier customers, like Bell South and Bell Canada, are using the Centrex IP solution in proof-of-concept trials with some of their biggest Centrex customers, who are viewing the service as part of their contract renewals. Gail Young, director of Nortel's mobile office unit, says these end customers are eager to begin implementing VoIP, but plan to do so in an incremental manner, and will proceed with a mixture of ISDN-based Centrex and Centrex IP lines in the near term.
"Customers like the fact that they can mix and match regular and IP lines, and add advanced feature functionality on an incremental basis - for example, within a single department, or for a single group of users," says Young. "A key value proposition of Centrex has always been the ability to grow one line at a time. Centrex IP just extends that capability by letting you grow one IP line at a time."
Young also agrees that the Centrex market can benefit from some of the uncertainty surrounding the future of PBXs right now, especially as IP access makes Centrex offerings more flexible. In addition to simplifying moves, adds, and changes, Nortel stresses the ability of Centrex IP to extend service to remote offices and telecommuters, many of whom were not previously able to get features like integrated dialing along with the rest of their organizations. The distributed nature of IP networks, Young says, will help to address a lot of the negative perception of Centrex as being restricted by a particular central office service area.
Like Lucent, Nortel plans eventually to separate out its class 5 features from the underlying switch hardware, and adopt an software-based IP call control architecture. At the same time, the company is working to build a subset of these features on top of a session initiation protocol (SIP) stack, to allow for their deployment in native SIP environments. Aimed at more aggressive customers, this type of solution would supply basic Centrex functionality, but focus more directly on rapid development of new applications.
Centrex From Scratch
For both Lucent and Nortel, an open, decoupled environment for creating network voice applications is the end goal of their IP Centrex initiatives. For some much younger switch and service vendors with no legacy architecture to preserve, however, it's just the starting point.
These vendors, who mostly come from the softswitch and application server markets, have no problem putting together the IP interfaces, open APIs, and other components that let carriers build and deploy hosted voice services on broadband networks. They see their challenge, rather, as figuring out how to include enough of the features and functionality of a class 5 switch in their products to match existing levels of service, while maintaining exponentially lower price points than legacy products. And as any softswitch maker will tell you, this shouldn't mean re-inventing the wheel.
For the most part, softswitches have identified a fractional subset of "useful" or "relevant" class 5 features that they are concentrating on building. Similarly, they view IP Centrex as tasked with determining which of the current Centrex features users actually need, and moving as quickly as possible toward developing the new services they want.
"Eventually we've got to come up with a better name for what we're providing than IP Centrex, because IP Centrex has too limiting a definition," comments Bill Leslie, CTO of Longboard, Inc. (San Jose, CA - 408-571-3300). "Most people imagine that you're just taking the existing Centrex features and building IP versions of them. What we're really trying to do with our offerings in this space is build a set of features that enables group communications."
For Longboard, "group communications" means making Centrex more inclusive - extending features across a VPN, for example - as well as putting more control into the user's hands. Being able to control features, even the basics like call transfer and forward, through a web interface as opposed to a keypad is an important first step in this regard, says Leslie. But it's also crucial for Centrex providers to start combining these basic features with new capabilities like presence management, which let a user determine how and when he or she wants to be contacted, and to segment his or her inbound and outbound communications according to various priorities and preferences. Even standard Centrex features, notes Leslie, should be reassessed in light of new communications media and new user interface options, and often can be made more efficient.
"There are some features in regular Centrex that it just doesn't make a lot of sense to replicate using existing mechanisms. For example, the busy lamp field: There's a perfectly good way to do this in the IP Centrex environment, using Presence. Or intercom paging, which you could replace with a broadcast instant message."
Longboard is integrating these applications and others in its own product, the Longboard Multi-engine Application Platform (LMAP). LMAP provides the software infrastructure for service creation and service delivery, but relies on a third-party softswitch to do call signaling and protocol translation. As a platform for IP Centrex, Longboard represents a significantly different paradigm from embedding features inside the switch itself.
Sylantro (Campbell, CA - 408-626-2300) holds a similar perspective to Longboard's on IP Centrex, though their product differs slightly. Sylantro includes the softswitch itself as part of its offering, which it positions overall as a class 5 alternative. At the same time, the application servers are modular, and plug into the softswitch through open interfaces, so they could work with a third-party call agent as well. Sylantro's approach to services is a combination of prepackaged applications, which include class 5 features as well as enhanced services like find-me/follow-me and wireless integration. All features are made accessible through a web-based "portal," which can serve as the user's primary telephony interface, working in conjunction with the phone itself.
One of Sylantro's first customers, GoBeam, is using the platform to offer a kind of enhanced Centrex service, all delivered over IP. GoBeam's service is aimed at small- to medium-sized businesses, the type of customers who would ordinarily be prime candidates for a key system or, now, for an IP-PBX. But the combination of Sylantro's web-integrated apps with a creative pricing scheme (based on service bundles, rather than a per-line charge) seems to be winning the company a number of early subscribers.
Coming at the problem from a slightly different angle is media gateway vendor Sonus (Westford, MA - 978-692-8999), who recently purchased softswitch company Telecom Technologies, Inc. (Richardson, TX - 972-301-4900). What Sonus offers could be seen as more of a complement than a competitor to Longboard or Sylantro. The company markets its product as a class 5 switch replacement, but does not intend to supply on its own all of the features and applications it expects customers to demand. Where Sonus basically draws the line between features it will develop and those for which it will look to third-party partners is in distinguishing between minimum requirements and enhanced services.
"Today, we don't have the whole class 5 and Centrex feature set," says Rob Held, director of business development for Sonus. "So what we're doing to address customer opportunities is using our partners to fill the gaps. But we're also very clear with our partners that if in two years they're still only offering the same basic features, they will no longer be adding any value to our own platform."
Broadband Office (Falls Church, VA - 703-641-6000), an in-building LEC (BLEC) also known as BBO, who is Sonus' first local exchange customer, is planning to roll out something along the lines of an IP Centrex service to its subscribers using the Sonus platform along with a variety of third-party application servers. BBO's business model, which involves leasing bundled voice and data services as well as hosted software to customers in multi-tenant unit office space, is another example of how IP Centrex can support a broader range of vertical markets and combine with other hosted services more easily than its predecessor.
Extending The Premise
There's no doubt that when the environment shifts to IP, Centrex services will be on more of an equal footing with their PBX counterparts. But increased competition between hosted services and premise-based phone systems isn't the only result of this shift. The interesting question, as PBXs become IP-based call servers, is whether there is any distinction at all between a Centrex platform and a private enterprise phone system.
Of course, there will be more than one model in the marketplace. But increasingly, we're starting to see companies who started out promoting their products as IP-PBXs now moving quickly into the hosted telephony, or IP Centrex, space. This includes players such as 3Com (Santa Clara, CA - 408-326-5000), who has announced plans to bring its NBX architecture into the service provider environment, as well as Cisco (San Jose, CA - 408-526-7208), who is working on scaling its Call Manager system and enabling it to support partitioned subscribers. A number of smaller companies are entering this market, as well, and either concentrating exclusively on the "hosted PBX" concept, or marketing to both enterprise and service provider customers.
The idea of remotely hosted PBXs, which can serve multiple different subscribers on a common platform, builds on the fact that services can be fundamentally decoupled from switching and transport in an IP telephony environment. In fact, you can think of the "PBX" in this scenario as simply another standards-based feature server, which interacts with a more centrally located softswitch for call signaling. While the IP-PBX and softswitch could be co-located in the same facility, a more likely deployment would have the PBX servers located closer to the customer premises (in a data center, or even the basement of a multi-tenant building), with one softswitch controlling multiple PBXs.
This is the kind of architecture Tundo (Westborough, MA - 508-836-4333) imagines for its Network Telephony System (NTS). Tundo also sells its system as a more conventional, premise-based PBX, but thinks that IP Centrex, or telephony ASP, services are where the biggest opportunities for the future lie.
The reason old-style Centrex services were never able to significantly expand their customer base, says Ron Shilon, Tundo's vice president of global marketing, is that CO switches always lagged behind PBXs in terms of functionality: For example, they could never successfully implement CTI or call center applications. Shilon argues that the same will be true of softswitches. They will never be able, on their own, to implement all the features that customers want, like unified messaging, multi-party conferencing, etc. The difference is, with IP, there is another option: Leave the features to vendors like Tundo, while employing the softswitch to handle session control, protocol mediation, SS7 integration, and other network functions which it makes sense to handle centrally. Because the system is set up for remote, web-based management anyway, end users don't lose any of the control they would have if the servers were sitting on-premises.
While Tundo doesn't shun IP Centrex as a description for the type of service its platform can support, Shilon says it's the growth of ASPs in the data world that are doing the most to help build the market for hosted telephony. "Data ASPs have changed the model," he says, "by making hosted services for other mission critical applications a familiar concept. Voice was always black magic for IT departments, but today, IT people are getting less and less afraid of touching voice. And if they're already used to outsourcing other applications, then there's no reason why they won't be motivated to do the same with voice."
Netergy Networks (Santa Clara, CA - 408-727-1885), have for the past several months been working to transition their product from more of an enterprise-focused platform to one that could integrate with the rest of a carrier's network infrastructure (in terms of scale, back office interfaces, etc.). Netergy makes it clear that, in their perspective, the balance of power between Centrex and PBXs has shifted. "There is no longer any clear distinction between IP Centrex and the IP-PBX," says Cyrille Thilloy, Netergy CTO. "It's become more of a packaging issue, a matter of how you want to define the customer community you are targeting with a particular service."
Ultimately, Netergy sees the only real distinction as being between different types of hosted models. A service provider could choose, for example, to locate the PBX features alongside the softswitch and serve them out centrally, or they could go with a more distributed option, where localized call agents would be responsible for single subscribers or smaller groups. In the former situation, Netergy would work with a third-party softswitch partner; in the latter, they provide their own call agent. According to Thilloy, the features and basic platform are in place for providers to begin offering IP Centrex services today. The only major developments remaining, he says, are to standardize a feature control protocol, which would allow a softswitch to communicate with an independent feature server (today there is work being done around SIP to achieve this). Netergy is already offering a Java-based, open programming environment to third parties who want to develop enhanced services on top of its platform.
VocalData (Richardson, TX - 972-354-2100) is another vendor whose product is difficult to identify as either a class 5 replacement or an IP-PBX. In fact, VocalData was one of the first companies we knew of to explicitly target the IP Centrex market when it released its Voice Over IP SoftSwitch (VOISS) last year. Like Netergy, the VocalData system has both its own call agent and feature servers, but the company makes clear that its emphasis is on the latter, and is also open to partnerships with the more CO-style softswitch vendors. Among the differences between the VOISS system and an enterprise IP-PBX are the fact that VocalData's system is built on UNIX, and set up to run in redundant configurations, as well as its native ability to handle partitioned users. In addition to the standard set of PBX features, VocalData supplies its own IP-based, multiparty conference bridge and its own unified messaging platform.
Early Movers
Most businesses won't be able to go out today and find a provider who's selling IP Centrex in their area. A few ambitious startups are signing up customers in limited rollouts (see Ellen Muraskin's sidebar on GoBeam, for example), but most are still in various stages of beta trials.
What we are starting to see, however, is a number of service providers offering IP Centrex and other applications bundled around it on an indirect basis, through channels that could include CLECs, DLECs, ISPs, ASPs, or even VARs and interconnects. Essentially, these players, most of whom call themselves communications ASPs, have introduced another layer into the value chain for Centrex services, and are enabling companies to offer the service who don't themselves own any aspect of the telephony infrastructure. If their business models are successful, the result could be an immense proliferation of competitive Centrex offerings, reaching a scale and level of diversity that would be far greater than what has emerged from the competition between ILECs and CLECs.
Before, what guaranteed the uniqueness of Centrex was the fact of its being inextricably bound to a single, very expensive hardware and software platform. Now, Centrex is becoming just one application among many in the network, and runs off of servers that cost a miniscule fraction of the price of a class 5 switch. The result is providers like I-Link (Draper, UT - 801-576-5000), who maintain their own network of softswitches, gateways, and routers, and sell access to the platform via open APIs to third party application developers. I-Link hosts these applications, while their partners market the service to end users. The range of possible services is open-ended, but I-Link considers IP Centrex one of its biggest areas of potential growth moving forward, and says it already has at least a few developers writing Centrex apps to its platform.
According to John Edwards, I-Link's chairman and CEO, two key factors differentiate an IP Centrex application developed on top of I-Link from a more conventional Centrex. One is manageability: With an XML-based architecture, plus web and speech rec front-end interfaces, it is possible to give more control to the user without sacrificing the security or functionality of the application (i.e., you don't have to worry that the customer will break the service just by touching it). Tiered administration levels give users the flexibility to customize and make changes to their service remotely, while the core system management is left in the hands of the service provider. The second is pricing. Because I-Link switches all of their calls in software, they no longer have to charge on a per-port basis for access to their switch. Their partners, in turn, can offer different and more attractive pricing models to their end customers, as opposed to the standard monthly charge per line.
"When everything to you is a port or a minute," says Edwards, "you're pretty limited in terms of how you can price the service. But when you move into a model like ours, where everything is software based, and running on off-the-shelf servers, I don't have to think of my ports and my equipment as part of the same resource. So even if you keep the same pricing structure, the price point becomes substantially lower. But you're also able to change the pricing altogether. We try to give our customers more flat-rate, packet-oriented schemes in order to encourage them to adopt more flexible pricing themselves."
Other providers are taking a similar approach to I-Link's in how they price and sell their service through partners, but are developing the applications - or at least the basic IP Centrex feature set - themselves. NetCentrex (San Jose, CA - 888-547-9897), for example, is working on bundling an IP Centrex application with access to its network of softswitches and voice-over-IP gateways. Despite its name, NetCentrex started out with more of a consumer-focused service, though it has always been in the company's plan to target enterprises as well.
Driving its efforts forward is an agreement with major carrier France Telecom, who plans to use the NetCentrex platform to offer IP Centrex services to its own corporate customers. NetCentrex is able to build upon some of the basic class 5 features it developed for the residential market, as well as the media processing technology it gained through its recent acquisition of MG2, a call center systems company based in Paris. MG2, for instance, had its own IVR platform, which NetCentrex will integrate as part of the IP Centrex application. While the company's current call-control architecture is based on H.323 signaling, their plan is to migrate over time to support MGCP, controlling features centrally in the network and communicating through gateways to the client endpoints.
TalkingNets (Wilmington, NC - 910-332-1800) is another player offering to host IP Centrex, as well as other business telephony applications, for a variety of channel partners. The company launched its first commercially available services in Denver and Cleveland late last year (by the second part of 2001, they expect to be rolling out around four new markets per month). Its service provider customers for those markets are a regional ISP and a BLEC, respectively, and TalkingNets supplies both the basic local dialtone as well as an enhanced IP Centrex, or what it calls "virtual PBX," service.
Unlike I-Link, who built the majority of its platform on its own, TalkingNets is using the telecom technologies (now part of Sonus) softswitch, and an application server provided by Broadsoft (Gaithersburg, MD - 301-977-9440), now part of Unisphere Networks). The Broadsoft server comes pre-integrated with some class 5 and enhanced features (accessible by the end user through a browser interface), and also supplies a set of open interfaces and Java-based programming tools to create new services. TalkingNets plans to include unified messaging/communications, multiparty conferencing, and other services that it and its partners will develop.
One thing that sets TalkingNets apart is its early adoption of SIP phones. The company has built its service on SIP, hooking up to Level 3's network of SIP gateways in the backbone, and stressing the importance of distributed intelligence at the endpoints. While hesitating to declare any of the generally available SIP phones to be fully mature, TalkingNets vice president of marketing Mark Cortner says that the company has chosen Cisco's 7960 model for its initial offering, and is working out deals to provide the phones at no extra charge to customers. Having an intelligent IP client device like a SIP phone, Cortner says, solidifies the value proposition to customers, as well as making things "as simple as possible for the operations team, since they don't have to deal with different installation models." In the future, he adds, the device will also be an important interface to services like one-touch conferencing.
Congruency (Rochelle Park, NJ - 201-712-5590) shares a similar feeling about the importance of IP endpoints, yet at the time it began building out their network for hosted telephony services, in 1998, couldn't find any third-party products to satisfy its needs. Having a clear idea of its requirements, though, the company decided to build its own. As a result, congruency now offers its service provider customers a full-featured, H.323-based IP phone, the i.Picasso 6000, along with the call switching and enhanced services that they host in their network. As third-party IP phones become available (which we think they will in the next year, from vendors like 3Com, Pingtel, and Siemens), congruency plans to certify and offer those as well, starting with H.323 and moving over time to SIP.
"Our belief has always been that if you're going to build a next-generation network, it shouldn't just provide me-too services," says Ralph Hayon, congruency's CEO. "We developed the i.Picasso phone as a delivery mechanism for the enhanced services which we saw as the real strength of our network. It had to be a device that had the look and feel of a traditional telephone, but also allowed us to present things visually to the user, and to change the user's experience with enhanced services."
congruency's development efforts did not stop at the phone: In fact, congruency has built the majority of its network infrastructure on its own, including gatekeepers, application servers, and applications themselves like voicemail, auto attendant, and a complete set of IP Centrex features. All of it runs on Compaq servers and Linux, hosted at Exodus facilities, and connects to the network through Cisco switches and routers. Like other communications ASPs, congruency intends its service to be delivered over a broadband pipe (most likely either DSL or a T-1) to the customer premise. Hayon says that the company has had its most success to date working with data CLECs and ISPs as its channel partners, but adds that he is also starting to see interest from interconnects and telephony VARs. With the benefit of past experience in supporting voice systems, the latter can now begin to resell IP Centrex services on an ASP model as an alternative to their more traditional CPE business.
As the year progresses, we expect to see more ASPs offering IP Centrex as a hosted application for channel partners. At the same time, we think the channel itself will continue to grow and diversify, pulling in players from CTI, the web, and broadband networking. Each of these will have the opportunity to customize and add value to the service based on their own core competencies, along with their creativity. And whether or not it's marketed as IP Centrex, chances are that by the end of 2001 you'll have at least a couple of options for getting hosted, PBX-style features and more from the new IP network.
Bundling Possibilities Of IP Centrex
We got a good look at the bundling possibilities of IP Centrex from a startup called GoBeam (Sunnyvale, CA - 408-991-0980). GoBeam seems to have learned a lot of lessons from the market's previous pitfalls. Their offering, pitched at enterprises in the seven-to-75-employee range, relies on VoIP between their Ethernet "Client Adapters" on the CPE side and a Sylantro (or soon) Vovida softswitch in their service office.
The service only starts with outsourced switching. Throw in find-me/follow-me, a thousand monthly minutes per seat of domestic calling, auto attendant, DID, a configurable choice of call treatments per caller category, unified messaging, drag-and-drop conferencing, application sharing, speed dial, and - I especially like this one - integrated instant messaging. And make the whole thing a web application, so that you can reset the follow-me or pick up your messages anywhere you touch down. Make it voice-accessible, too, so you can reset find-me/follow-me, and hear voicemail from any phone.
One of the first things GoBeam has learned is to leave the broadband business itself to others. They partner with data LECs to run their packets across a T-1 (or a DSL, cable, fixed wireless, or frame relay connection) from client adapter to mux device and thence ATM backbone. From there, the traffic gets sent via DS3 to their service office, now in San Ramon, CA. "It never touches the Internet," Jeff Stern, VP and co-founder, points out. At the moment, the company isn't placing its bets on any particular broadband channel. "All I need is Ethernet on one end and ATM on the other," says Stern.
They've also learned that (1) most users will not be deserting their dial pads for mice and keyboards anytime soon, and that (2) GoBeam cannot, on their own, amass the staff necessary to grow and support this business the way it must be grown and supported. So they're partnering with Toshiba interconnects - so far, ATS Communications, in Concord, CA, and Olympic Systems, in Sacramento. The desktop phone connecting (via Amphenol cable) to the client adapter is a Toshiba; the client adapter hooks up 24 lines, simulates ring and dialtone, captures DTMF, does a-d conversion and packetization, and sends MGCP signaling to the softswitch. The Sylantro Softswitch speaks fluent Toshiba, and so those users faithful and familiar with their digital handsets can operate as before. Analog phones can also plug into the client adapter, and use traditional hook-flash to conference and transfer.
The beauty part comes to those willing to use a browser GUI for click-to-dial-back, unified messaging, and to at least configure - if not operate - their phones. GoBeam has designed a user-doting interface, which they call their Dashboard. It seems to put to use all the lessons learned in human factors engineering. On the left-most panel, classify your routine callers, configure their greetings, and tell the switch when, where, and if they can reach you. Right of that panel is a directory/speed-dial list; part of GoBeam's service is populating this with a company's employee roster and customer list. One panel over, a conferencing window has slots for five names to be dragged from the directory or hand-entered phone numbers. Each one can be muted, and these types of audioconferences just consume basic package minutes. Although pricing policy is preliminary, GoBeam will probably throw in half an hour per month of enhanced audioconferencing, including T.120 application sharing and recording, and an unlimited number of participants. Time beyond 30 minutes will be charged separately.
The rightmost panel holds your instant messaging - one window for buddy list, one for chat, one for entry. If anything is going to break IM out of its teenage pigeonhole into popular business usage, this will: the quieter (and further-reaching) equivalent of a shout over the cubicle walls. The best part is that it sits on the Dashboard with all the other tools, so its use will be encouraged. There's also a link on the Dashboard to e-mail; GoBeam assumes that we're already partial to a particular e-mail program and doesn't do that part over. A call log keeps tabs on incoming and outgoing calls, all clickable for dialing. You can also use the GUI to program your Toshiba phone buttons for speed dialing, call park, call pick-up, and the like.
I was able to demo this system in a backwards sort of way with a login to GoBeam's demo phone. I browsed over, logged in, and entered my own home-office phone number to be dialed. My phone rang immediately. When I picked up, I heard a lot of background office noise. I called "hello, hello," until Stern, passing by, picked up.
In entering my number on the web page, I had ordered the switch to dial me on the PSTN through its gateway. The switch, programmed to assume that anyone who dials "out" is at his extension, puts the Toshiba in Sunnyvale on speaker and takes it off hook. That's why I, experimenting with the GUI without notifying GoBeam, found myself talking to an unattended but off-hook Toshiba station set.
Stern and I took the opportunity to discuss GoBeam some more, and to sample a voice connection that matched the PSTN's on its best day.
GoBeam, of course, is happy to re-label Dashboard for resellers and customers; Interconnect ATS is going this route. GoBeam's game plan includes three more Service Offices in other parts of the U.S. In Northern California, they have 60,000 telephone numbers to assign, and "can get more," says Rob Stevenson, CEO. They're also launching a drive to recruit more Toshiba interconnects.
For now, the service office platform combines a Sylantro switch with a conference server from Latitude Communications, a unified messaging server from Centigram, OSS Server, billing engine from Portal Software, a Lucent media gateway, and an Oracle database. To be added is an LDAP server and a SIP-speaking Vovida softswitch. The Sylantro uses MGCP signaling to the Lucent media gateway; a future Vovida softswitch will work with a Cisco-based gateway using SIP. SS7 will be hooked up early in 2001; in the meantime, the Sylantro switch relies on a PRI interface.
As a voice ASP offering bundled with basic phone service and switching, GoBeam may succeed where separate UM or personal assistant pieces have received only lukewarm acceptance. Dave Laukat of ecFoods (see Muraskin File, this issue), a pilot customer, doesn't use half of the system's bell and whistles, but really likes the ease of message answering and resetting find-me/follow-me on the browser-based dashboard. Webley and Portico also offered browser-based (and phone-based) routing, screening and messaging features (in addition to dial-from-anywhere features GoBeam currently does not provide), but they don't have these VoIP and switching cost savings to offer, local-number portability, or an interconnect partner's familiar face. With these, GoBeam may find it easier to get in the door.
Pilot pricing, not yet settled, is on cellular model: a retail $49.95 per month per desk for 1,000 minutes of calling, a 30-minute voicemail store, and 2MB of fax mail. Extra voicemail minutes, meeting minutes, fax mail minutes, and (maybe) videoconferencing are planned as add-on packages. According to GoBeam's calculations, that's about a 50% savings per desktop over plain phone service alone.
-Ellen Muraskin
