What a difference a year makes. No matter how you slice and dice the numbers, the popularity of voice over IP grew by leaps and bounds in 2001. The number of installed voice-over-IP networks, the number of players in the voice-over-IP arena, the dollars spent on voice-over-IP products, the number of channels shipped and even the capacity of voice-over-IP products -- all of these elements more than doubled last year.
A few years ago, voice over IP was the domain of just a handful of early pioneers including 3Com Corp., Cisco Systems Inc., Clarent Corp., Nuera Communications Inc. and Hypercom Corp. But now this convergence technology is finally being embraced by more traditional networking and telecommunications vendors that previously viewed voice over IP as a serious threat to their installed base.
According to an ongoing survey conducted by Mier Communications Inc., these early voice-over-IP pioneers have been joined by the classic PBX vendors -- Alcatel SA, Avaya Inc., L.M. Ericsson Telephone Co., Mitel Telecommunications Systems Inc., NEC Corp., Nortel Networks Corp. and Siemens AG. Within the last year, all of the vendors have introduced viable voice-over-IP products, often in the form of add-ons, which "IP-enable" the latest versions of their time-division multiplexer (TDM) and switching matrix-based PBXs. We have seen this technology working - with good to excellent voice quality and acceptable to very good reliability. In the early days, that wasn't always the case.
Interoperability among voice-over-IP products has been a major stumbling block to widespread acceptance of the technology. The ITU-T's H.323 "umbrella" standard, the first posed for voice-over-IP interoperability, proved complex and difficult to implement. As a result, other less-unwieldy standards were posed in its place, and until recently, we have seen little consensus on which voice-over-IP standards would be most widely implemented.
However, we are now beginning to see some general agreement within the vendor community about where voice-over-IP standards are headed. Based on Miercom's latest survey of voice-over-IP vendors, a number of relative standards -- among them the ITU-T's H.323, the Internet Engineering Task Force (IETF) Session Initiation Protocol (SIP) and Media Gateway Control Protocol (MGCP), and International Softswitch Consortium (ISC) specifications, and the ITU-T's H.248/ Megaco -- will all coexist. So don't expect any single standard to emerge as the basis for interoperability -- at least not anytime soon.
The prevailing opinion among vendor respondents is that H.323 will become the enterprise legacy standard, while MGCP and H.248/
Megaco will be used between carriers' call agents and other media gateways.
Many believe SIP will be used between call agents and between call agents and residential IP phones. In other words, how protocols are implemented will depend on where the voice-over-IP equipment is situated in the network. Clearly there will need to be coexistence among all these standards -- at least in the short term.
We first examined trends in the voice-over-IP arena with a survey asking vendors about their IP telephony products (www. nwfusion. com, DocFinder: 2633).
Miercom conducted a second survey late last year, which was sent to 175 voice-over-IP vendors. All told we ended up with 84 valid responses from 79 vendors. In our estimation, this survey response base represents about two-thirds -- from about 65 percent to 70 percent -- of the total universe of vendors who make and sell voice-over-IP equipment to end users, according to Miercom's analysis of the market.
In summarizing this voice-over-IP data, we've learned there are new product categories emerging, capacities are on the rise, prices are coming down at the high end, the standards arena is shaking out and vendors are forming interoperability alliances.
Trend no. 1: New product classes hit the market
Product categories resulting from our survey comprise complete voice-over-IP systems, voice-over-IP gateways, voice-over-IP gatekeepers, IP PBXs, public switched telephone networks (PSTN) central office replacement switches, IP phones and "other".
Interesting trends emerge when you compare this 2000 data with our previous results tracked in 1999.
First, the voice-over-IP product category that has seen the most rapid growth is IP telephony, which includes IP PBXs and IP-enabled PBXs. The number of discrete products and IP-telephony systems, which provide PBX-type functions over an IP transport and thus are candidates for replacing the traditional circuit-switched PBXs, has more than doubled from year-end 1999 to year-end 2000.
Just a year ago, the IP PBX segment belonged to early adopters such as 3Com and Vertical Networks Inc., which targeted the "100 lines and under" market. The technology was being deployed in small companies and experimentally in larger ones, but as the promises of IP PBX technology (such as unified messaging and "follow me" phone services) were more widely recognized, the traditional PBX vendors began to add IP capability to their traditional systems to maintain their installed bases. This adoption by more entrenched PBX vendors has added a sense of legitimacy to this market segment.
As one vendor says, "It's not a matter now of if we go to voice over IP, but when." However, vendors that are "IP enabling" their PBXs are about six months behind vendors of "native" IP PBXs and have some catching up to do.
Our research turned up a total of 25 shipping IP telephony systems. Native IP or IP-centric systems on the market now are Avaya's new IP600, 3Com's NBX 100, Alcatel's OmniPCX 440, Cisco's AVVID/CallManager, Shoreline Communications Inc.'s Communications System, Siemens Enterprise Networks' HiPath 5300/5500, Sphere Communications Inc.'s Sphericall and Vertical Networks' InstantOffice.
The more traditional PBX vendors that offer IP-enabled versions of their erstwhile TDM- and switching-matrix-based PBXs include: Avaya with IP add-ons for the Definity; Mitel, with its Ipera 2000 evolution from the SX-2000; NEC America, with its NEAX 2400 IPX; and Nortel, with add-ons for the Meridian 1 and the Business Communications System, which was the Norstar system.
Our research indicates there isn't widespread adoption of IP-enabled PBXs because the technology is fairly untested. So corporations are deploying native IP PBX systems, such as Cisco's AVVID, experimentally for specific applications or within departments or branch offices, while maintaining their traditional PBXs corporatewide. Furthermore, full-scale deployment of IP PBXs has been hampered because these packages do not offer the array of voice features available on traditional PBXs. However, some IP PBX vendors -- Cisco, for example -- are rapidly adding new phone features to their products to compete more aggressively against the traditional PBXs.
Another notable voice-over-IP product trend is the growing number of complete voice-over-IP systems now offered. These products incorporate one or more voice-over-IP gateway models, along with a voice-over-IP gatekeeper or other high-level voice-over-IP call-control system (also often called a softswitch or call agent), and a management system. As recently as 1999, most vendors offered only piecemeal or partial voice-over-IP systems. With the complete systems, the buyer benefits by having a system that doesn't include a lot of third-party components, which might have been sold as fully interoperable, but may not have been. There is the advantage of working with one vendor/supplier as opposed to many.
Products that fall into our "other" category include software applications for billing and accounting, settlement services that monitor and report transactions for charging back voice-over-IP services within the network, and gateway and gatekeeper/softswitch software, which is typically, though not exclusively, sold to systems integrators and OEMs.
One unique product in the "other" segment is RAD Data Communications' IPmux-4, which provides an unusual form of a voice-over-IP gateway. The product delivers TDM over IP.
An IPmux-4 at each end of an IP network translates T-1 frames and signaling into IP packets, transmits these over an IP network, and then decodes and reassembles the complete T-1 signal at the other side of the IP network. As with classical T-1 equipment, the IPmux-4 accepts voice, data and video inputs, allowing true convergence at a cost of about $70 per equivalent DS-0-type channel.
Trend no. 2: Capacities on the rise
We asked vendors of gateways, gatekeepers/softswitches, and PSTN/central office replacement switches to delineate their products' maximum capacity in terms of concurrent channels supported (See "VoIP standards support, current and future," page 64). They detailed maximum configuration for a single-gateway chassis and multiple gateways that could be logically combined -- within an H.323 zone or MGCP domain. We also asked whether these products were targeted for enterprise or carrier/ISP environments.
The average maximum voice-over-IP channel density on a single-gateway chassis targeted to enterprise networks is 366 (based on 15 discrete systems). This represents a range of channel densities, from only 23 channels on Siemens' RG2500 up to 1,920 maximum on Clarent's Backbone High Density gateway. This average maximum represents a tripling of capacity in 1999 when the average maximum was just 100 channels per single-chassis gateway. Our research shows that about 90 percent of the phone systems installed in enterprise applications are 100 lines or less. So a tripling of capacity clearly addresses that part of the market. However, the remaining 10 percent of enterprise applications are way beyond that -- up to 15,000 lines and beyond are often in distributed applications. Many of the IP-enabled PBXs are addressing that higher end of the market, where they are more widely installed already.
On products targeted to carrier and ISP environments (15 discrete systems were included), the average maximum channel density is far higher -- 1,989 on average. This represents a range from 24 channels on Cisco's 3810, up to 15,000 on Lucents' Exchange Plus PSTN/central office replacement switch. This, too, represents more than a doubling of channel capacities from 1999. It's also notable that systems targeted primarily to the carrier market were not yet widely available in 1999.
We also asked vendors about their systems' ability to scale multiple gateways within an H.323 zone, which means they are under the control of the same voice-over-IP gatekeeper, or within a MGCP-based domain, meaning they are under control of the same call agent or softswitch. Enterprise products support 3,683 channels per zone or domain, while carrier systems average 10,950 channels.
Several vendors -- Lucent, Siemens and Sonus Networks among them -- say their gatekeeper (or softswitches) could handle an unlimited number of channels, noting that there are no architectural limits to the maximum number of concurrent channels that can be managed or aggregated. Vendors noted, however, that memory, CPU constraints or other configuration issues limit even these systems.
Trend no. 3: Pricing fluctuations
So how much will voice over IP cost in 2001? On average, between $500 and $600 per channel, according to our analysis. This is roughly the same as the per-channel price a year ago when prices ranged from about $500 to $700 per channel. There were more expensive products a year ago when the technology was newer, and they have since coalesced at a slightly lower price, which brings the top down. It's now a narrower range.
Our per-port pricing for shipping voice over IP was based on the following criteria:
- For voice-over-IP gateways and/or gatekeepers, the price includes two gateways plus one gatekeeper (or call agent), which would comprise a typical enterprise deployment, for a number of T-1 links that the vendor specified. We asked the vendors to ensure that the per-port pricing included all basic PBX features (such as call transfer, call forwarding and conferencing) and capabilities including management, interactive voice response, two-stage dialing, autodetection of modem calls, retention of CDR records, and any supported redundancy.
- For PSTN/central office replacement switches, we requested a system price based on one switching system/chassis/rack, configured to support 20 T-1s (480 input voice/DS-0 channels). We also asked for a per-channel price based on a fully expanded system supporting the maximum number of T-1s (which varied from vendor to vendor). Prices likewise were to include all basic features and capabilities. This class of products is generally sold only to service providers. With the proliferation of ISPs and competitive local exchange carriers, we found it difficult to make a comparison with the way things were a year or more ago.
We asked IP telephony (IP PBX) vendors to provide per-station pricing based on a 100-station configuration that could be all IP phones, all analog phones or a 50/50 mix (if both are supported), with one T-1 trunk to a central office or 24 foreign exchange office analog ports (if T-1 was not supported). This price was to include full voice-over-IP support and management software.
As noted before, a 100-or-fewer line system is the typical size deployed in 90 percent of enterprise applications. The 50/50 mix of phones is more theoretical. IP PBX configurations often can't be compared one for one with classical PBXs. For example, stations such as soft phones (software on PCs with a microphone) don't exist in the classical PBX environment. In the IP PBX world, there tends to be more than one station per employee (typically an IP phone and a soft phone, both of which are IP PBX stations). There is seldom more than one station per employee with classical PBXs.
The average per-channel price for voice-over-IP systems (including PSTN/central office replacement switches) was $571. This represents a fairly wide range of prices from a low of $100 per channel (for RAD Data Communications' IPmux-4) to a high of $1,550 per port on Memotec's CX800 system. As it turned out, the median price (around $500) among all voice-over-IP systems included in this analysis virtually matched the average price.
The average per-station price of an IP PBX is $514. This is based on an averaging of 13 systems, ranging from a low per-port price of $331 with Vertical Networks' InstantOffice to a high of about $790 on Alcatel's OmniPCX. It was noted that PBX topologies can vary significantly, especially with so many different station options.
It is noteworthy that more than half of the vendors didn't provide detailed or configuration pricing. Many vendors -- particularly those at the high-capacity end of the market -- don't want to go public with pricing, citing the old line: "It's heavily configuration-dependent." Translation: prices are heavily discounted to major accounts, so the vendors don't want to reveal prices.
Trend no. 4: Peaceful coexistence in the standards arena?
In 1999 many vendors -- speaking mostly off the record -- thought the voice-over-IP standards arena was in a state of utter chaos and would remain so for at least several more years.
At issue was the fact that the most widely embraced standard at that time -- the ITU-T's H.323 umbrella standard -- was far too complex to implement efficiently or interoperably. Originally designed as an end-to-end communications standard for videoconferencing over packet networks, H.323 was retrofitted for voice-over-IP applications. The result was a standard that defines far more functionality than is necessary for most voice-over-IP environments.
H.323 also allows for many options in implementing it, which means vendors interpret the standard in many ways. While flexibility is typically a good thing, it can be problematic where interoperability is concerned.
Also, various incompatible versions of the H.323 standard exist. Version 1, which is not forward-compatible with the most widely supported Version 2; and Versions 3 and 4. Version 3 is a ratified standard at this point but Version 4 was not as we went to press (see graphic, right). Clearly so many versions of the same standard have contributed to industry confusion.
Confusion or not, H.323's Version 2 is the most widely implemented today, present in 49 products out of the 84 we analyzed. However, when we looked at what vendors are planning to do in the standards arena, other standards are gaining some major ground.
These include MGCP, a protocol that addresses control of media gateways, but it does not -- as H.323 does -- specify a complete end-to-end communication. Vendors supporting or planning to support MGCP within the next six months include Alcatel, Avaya, Cisco, Clarent, Dialogic, ECI Telecom, Inter-Tel, Motorola, Netcentrex, Syndeo, Telecom Technologies, Telogy, Unisphere Networks, Vertical Networks and Vsys.
Two variants of MGCP exist: one, per RFC 2705, under control of the IETF, and another version under the auspices of the ISC. Among the 36 products supporting MGCP, 22 support the RFC 2705 specification, while nine specifically support the ISC version.
H.248/Megaco, a joint undertaking of the ITU-T and the IETF, combines elements of the IETF's MGCP with the ITU's H.323. The main thrust of Megaco is to permit greater scalability than allowed by H.323, and to address the technical requirements of multimedia conferencing. At present H.248/Megaco is not nearly as widely supported as H.323 or MGCP, but vendors say they plan to implement it on 30 products within the next six months. The vendors that support H.248/Megaco now or plan to within the next six months are Avaya, Alitgen Communications Inc., Alcatel, Cisco Systems Inc., Clarent, Dialogic Corp., Motorola Inc., Sonus Networks Inc., Unisphere Networks Inc., Vertical Networks and Vive Synergies.
Yet another standard gaining widespread acceptance is SIP. SIP is an application-layer signaling protocol that specifies call control for multiparty sessions, IP phone calls or multimedia distribution. Unlike H.323, which is based on binary-encoding, SIP is a text-based protocol that vendors tell us is much easier to implement.
The vendors that currently or plan to support SIP are Avaya, 3Com, Altigen, Alcatel, Cisco, Com2001, Ericsson, ECI Telecom, NEC, Nortel (on the Business Communications Manager), Nuera, Pingtel, Shoreline, Syndeo, Telecom Technologies, Telogy, VocalData and Vsys. SIP is supported on 24 products and is planned to be implemented on 39 more within the next six months. This will boost SIP's prominence significantly.
When we last examined the standards issue about eight months ago, half of all vendors said they were supporting H.323 in some way. That's still true now, but it is clear that competing standards are encroaching on H.323's territory.
From vendor interviews we conducted for this review, we gleaned a general consensus that by year-end 2002, most (more than 50 percent) voice-over-IP equipment will interoperate. However, this clearly will not be based on one standard but many.
Miercom will be testing these claims in our continuing voice-over-IP interoperability labs, the first of which will run at ComNet 2001 in Washington, D.C. this week.
Trend no. 5: Interoperating partnerships
If a voice-over-IP product is going to work with any other vendor's gear in general, that gear is going to be shipped by Cisco. Twenty-eight vendors say they interoperate with Cisco voice-over-IP products, with 10 reporting that interoperability was achieved specifically with Cisco's AS5300 gateway. That's not surprising, considering the prominence of Cisco's voice-over-IP equipment in enterprise and carrier networks.
The second most frequently cited voice-over-IP partner and product was Microsoft's NetMeeting. A few vendors also reported interoperability with Microsoft's Outlook e-mail application.
Lucent was the third most frequently mentioned partner (by six vendors on various products), although this involves equipment that has now largely been assumed by Avaya. Another reference product for achieving interoperability is RADVision's gatekeeper -- the fourth most commonly cited interoperable partner (by five vendors).
About half of the vendor respondents reported they had conducted widespread interoperability trials, but were generally not willing to go on the record with partners' names -- mainly citing nondisclosure agreements for the lack of specifics.
Our assessment of the thorny voice-over-IP interoperability issue is that while we're seeing more willingness among vendors to cite interoperable partners than a year ago, widespread interoperability in the voice-over-IP marketplace doesn't yet exist. Until this interoperability is proven in the public domain, all of the gains voice-over-IP products have made in the past year will be bridled by the fact that making an end-to-end call using this technology is probably not going to work.
This story, "Interoperability affects booming VoIP marketplace" was originally published by Network World.