IP telephony, while still relatively new, is poised to revolutionize corporate telecommunications by at last enabling users to haul voice and data traffic over a single wide area network. International Data Corporation, predicts that the Internet telephony market will grow from US $3. 5 million in 1995, to US $560 million 1999. Meanwhile, Tarifica in the UK estimates that AT&T will to lose between US $620 million and US $950 million in international calls to the Internet by 2001.
The problems with telephone systems as they are today, are getting increasingly higher for long distance phone class. Imagine even local ISP’s (Internet Service Providers) offering such low rates as $. 02 a minute phone calls from Rapid City to Sioux Falls. The idea behind IP telephony is enticingly simple, convert voice into packets for transmission over a company’s TCP/IP network. Although simple in principle, IP telephony presents some tough technical hurdles for vendors to surpass.
This report will provide readers with a comparative analysis of the IP telephony products currently on the market, or scheduled to ship that will help you understand the significance of incorporating this technology for the benefit of the consumers. The basic concept behind IP telephony is deceptively simple: convert voice into a series of packets, and transmit them across an IP network to be reassembled on the receiving end. While theoretically simple, IP telephony designers face a number of technical challenges, many of them similar to those encountered in designing a digital cellular network.
These challenges make designing a good IP telephony gateway a significant undertaking. Vendors have approached IP telephony from several different angles. Some vendors have leveraged existing hardware such as Dialogic’s multi-port voice cards in conjunction with their own software. Other vendors have built their own DSP cards designed specifically for this task. Other vendors have built stand alone network appliances. And one vendor has taken this a step further by creating an Ethernet IP telephone.
Vendors designing IP telephony products face two fundamental issues. One is latency, or in laymen’s terms the time required for data to travel from point A to point B. Latency does not in of itself destroy data, but it makes for awkward conversation when it climbs above a few tenths of a second. The other fundamental issue is packet loss. TCP/IP, in its present incarnation, provides no guarantees to users that they can transmit a given amount of data within a given period of time. Network performance can fluctuate from moment to moment.
Sometimes data will be transmitted immediately, sometimes it will be delayed or not sent at all. In many respects, the problems facing designers are similar to digital cellular network design in that the underlying transport medium provides no guarantee that data will arrive intact. Therefore, vendors must assume the underlying network is unreliable and compensate for this by using error correction and interpolation techniques to minimize and compensate for lost or severely delayed packets. These problems are not insurmountable by any means, but they do require serious attention.
The first IP telephony gateway products were based on industry standard voice cards such as the Dialogic series voice cards. These cards are designed to play and record announcements to callers, and are typically used in office voice mail or interactive voice response systems. These vendors had the benefit of being able to get their products to market quickly. However, these systems also have some important limitations: Industry standard voice cards are not designed for real-time applications like IP telephony.
The IP telephony software must deal with audio in discrete chunks rather than as a continuous stream. This introduces latency (delay) into a conversation. Industry standard voice cards are not designed to compress and decompress audio. This means the host PC containing the cards must do all of this processing on the host CPU. This also introduces latency, and limits the scalability of the gateway to no more than a few simultaneous conversations. Industry standard voice cards are generally not very easy to install and configure; they are not plug and play cards.
Many of these systems use proprietary protocols to deliver phone calls, which means you are locked into one vendor’s solution. At US $1000 to US $2000 per port, these solutions are on the expensive side. These gateway servers, while they served an important role in the market (i. e. they were the first working products to hit the market), have a limited future since they use hardware which was never designed for this kind of application. In order to remain viable, vendors using this design will need to shift to hardware that is designed specifically for IP telephony.
PC Based Server Gateways Based On Specialized DSP Cards Once the concept of the IP telephony gateway was proven, vendors sought to address the shortcomings of the existing voice cards by engineering cards specifically for IP telephony. This approach offers dramatic improvements in gateway performance, sound quality, and overall scalability of the service. By off-loading the dirty work of sampling, compressing, and transmitting audio from the host CPU, these cards reduce latency, improve audio quality, and allow a single server to handle more simultaneous conversations.
Several vendors have created DSP cards specifically designed for IP telephony: Dialogic – They are working on a next generation telephony card (known as DM3 IP Link) that is optimized for IP telephony applications. Third party software from vendors such as VocalTec will be required to implement a complete solution. Micom – They have a whole suite of analog, T1 and E1 cards designed specifically for IP telephony. They provide a turnkey solution that completely offloads processing work from the host CPU. Their solution works on DOS, Netware, and Windows NT latforms. Micom provides the software, so no third party products are required.
Natural Microsystems – They have developed a card called Fusion that is engineered specifically for IP telephony. This card is typically used by developers to create customized network telephony applications. It has also been incorporated into IP telephony gateways as an OEM product by vendors such as Inter-Tel. For customers desiring a PC based IP telephony gateway, this is definitely superior to solutions where real-time processing is done on the host CPU.
These systems will typically offer better sound quality (with fewer dropouts and less latency), more reliable operation (since the application is less OS dependent), and more scalable. Some of these solutions, in particular Micom, also come in at a lower per port price point (US $500 to US $1000 per port). Stand-Alone Network Appliances The next wave in IP telephony products will come in the form of stand-alone network appliances resembling Ethernet hubs. To this end, Cisco recently announced that they are adding H. 323 compliant IP voice gateway capability to their high-end router product line.
By adding a module to their routers, you will be able to combine internetworking and IP telephony services into a single box. A number of other vendors, including Shoreline Teleworks and Touchwave Communications, are working on hybrid switching products which, in effect, make PBXs obsolete. These products come in the form of boxes that resemble an Ethernet hub. Each box is connected via RJ14 wiring to outside phone lines and telephone extensions and via RJ45 to a company’s Ethernet network. Each hub is essentially a small PBX with voice mail/attendant, etc.
When the hubs are connected to a company’s IP network, these small systems talk to each other across the network and behave like one large system. In effect, what you have is a distributed PBX, which uses a company’s IP network to haul telephone calls between switching nodes. Other vendors are working on stand-alone devices, which will just do POTS to H. 323 conversion in an appliance form factor. This approach will make a lot of sense for small offices which do not require or want a dedicated server for POTS to IP conversion, and for larger companies wishing to replace their central PBX with many internetworked mini-PBXs.