The Lost World

Most of the telecom industry believes Internet protocol (IP) networks are the next step in communications evolution. But the wireless industry isn't quite ready to take that leap. It is betting on intelligent networks (INs), which could end up facing extinction in short order. If IP telephony overcomes its growing pains successfully, the IN has the potential of becoming nothing more than a stegosaurus.

When it comes to providing digital access, wireless service providers are on the front lines. But their backbone networks are another story. In telecom, the trend is to push intelligence out to the edges of networks. The only people who still believe in centralized intelligence -- typically in the form of "big iron" switches -- are wireless network operators and that great fossil collection, the regional Bell operating companies (RBOCs).

Instead of banking on centralized switching, however, wireless operators should be developing networks modeled after the Internet. This would allow them to make more efficient use of backhaul facilities. More important, it would position them to offer myriad new voice and data services.

One of the most intriguing new services is IP telephony. Invented by a small Israeli company in early 1995, IP telephony exploits the Internet's distance-insensitive pricing. When surfing the Net, it costs the same whether you are accessing a site down the block or a continent away. With IP telephony, the data just happens to be voice. Some believe IP telephony will take a big chunk out of international calling revenues.

Today, most IP telephony vendors see wireless as just another access medium. That's an attractive option in developing countries, where there isn't enough wireline infrastructure. But it does not take much imagination to see how IP telephony could benefit international travelers. Instead of placing expensive international phone calls, travelers will reach their home offices through local Internet gateways. With every call a local call, travelers will be more likely to make all calls from a single wireless handset -- enjoying the twin conveniences of mobility and consolidated billing.

There is no lack of synergy between wireless and IP networks. Both pioneer the use of low-bit-rate voice. Both promise new ways to integrate voice and data. And both offer development of new devices and services for end users.

GROWING PAINS It was only a decade ago that data, audio and video traveled over separate networks. Then the Internet came along. Initially, IP was designed to interconnect disparate data networks. It quickly became clear, however, that IP networks could handle a variety of data types. This led to the development of the World Wide Web.

The public switched telephone network (PSTN) employs circuit switching. Whenever you make a PSTN call, a dedicated circuit is established between you and the called party. You consume all of the circuit's bandwidth whether you are engaged in furious conversation or placed on silent hold.

The Internet uses a more efficient approach: packet switching. Because each packet contains its destination address, packets from different users may be intermixed freely over the same transmission facilities. Less bandwidth is wasted maintaining connections -- making more bandwidth available when you need it.

However, the Internet is basically a data network. Most people connect to it via personal computers. They primarily use it to access web sites (host computers) and exchange e-mail. The Internet has both advantages and disadvantages for handling real-time voice calls.

The main disadvantage for voice is that today's Internet doesn't guarantee the kind of real-time, synchronous, full-duplex communications we have come to expect from the PSTN. Instead, the Internet offers variable quality of service (QoS). Sometimes IP telephony rivals the PSTN in quality; other times it doesn't come close.

Three factors determine IP telephony's QoS: the access link speed at each end, the voice coding algorithm's bit rate and quality, and the current level of network congestion. When using the Internet, not much can be done about the third factor. But with a private IP network (intranet), there are several ways to improve the QoS.

Users who employ the Internet for international calls can expect to encounter transmission delays similar to those experienced over geosynchronous satellite links. Basically, the callers on each end must take turns speaking. (Techniques are being developed which would reduce these delays.)

Internet telephony advocates may have unrealistic expectations concerning how much delay and distortion people are willing to endure in order to save a few dollars; some claim callers naturally adapt to transmission delays. Are they fooling themselves? If wireless operators (both terrestrial and low earth orbit satellite) would band together and build their own global IP network, voice callers would not have to contend with Internet congestion.

IP telephony has other disadvantages. Not all PCs are equipped for IP telephony, and not all of those that are can interoperate. One solution to the first problem is to use IP telephony gateways that support phone-to-phone, phone-to-PC and PC-to-phone communications. Products such as Quicknet's Phonejack and PhoneNet's PacketPhone allow ordinary telephones to place and receive IP telephony calls.

Note the use of gateways complicates call set up. Phone-to-phone and PC-to-phone calls require connecting to a local gateway, entering the destination phone number and waiting for the local gateway to find another gateway local to the other party. There won't be widespread use of IP telephony until there are IP telephony gateways in most major cities around the globe. However, there are organizations such as Free World Dialup and Delta Three to encourage cooperation between IP telephony gateway service providers.

The second problem -- interoperability -- will be solved through the development and acceptance of standards. The International Telecommunications Union has established the H.323 standard for multimedia communications over packet-switched networks. The Voice Over IP Forum is promoting industry-wide interoperability as well.

In addition to reducing the cost of international calls, the Internet promises new features. For example, voice calls may be combined with video conferencing or document sharing. Instead of merely referring to a spreadsheet the other party may or may not have, IP telephony would let the user show the other party the spreadsheet. This even could be accomplished with wireless handsets by equipping them with special virtual displays, which Reflection Technology manufactures.

Microsoft's NetMeeting software combines IP telephony with application sharing. Two (or more) parties can talk, view and edit the same document at the same time. VocalTec makes another popular IP telephony software package called Internet Phone.

IP telephony will become even more attractive to wireless users as they embrace wireless data applications. It opens the door to integrated voice/data communications such as making and receiving calls while downloading e-mail. Furthermore, IP telephony will permit a digital cellular or PCS user to connect to a local gateway and complete multiple voice calls and/or data sessions without having to redial.

IT'S THE BITS Will wireless networks follow the Internet's lead, abandoning gargantuan switches in favor of desktop routers that simply send bits where they're told?

That is known as the stupid network thesis. It was first articulated by self-described "telephone company nerd" David Isenberg.

In his paper, "Rise of the Stupid Network," Isenberg argued that the RBOCs' IN strategy is based on a series of false assumptions. The RBOCs want us to believe voice will continue to account for most telecom traffic; that circuit-switched calls will predominate; and that we need centralized intelligence to help us share scarce bandwidth.

Furthermore, Isenberg disputed the claim that INs facilitate the introduction of new features. The IN model says that new features become available when the telephone company implements and releases them. In a stupid network such as the Internet, anyone can develop new features and distribute them via a web site. And any customer can buy, download and begin using such features immediately.

While wireless operators and RBOCs continue to deploy switches that look like mainframe computers, end users are purchasing phones that look more like PCs. There are now cellular phones with processors that can perform 70 million instructions per second.

In an IN, the network tells the data where to go. In a stupid network, the data tells the network where to send it. Giving control to end users is an important innovation, because ultimately it will give them the ability to control voice quality and add new features -- merely by upgrading device software.

But there is one potential limitation. How can a dumb network support mobility? There needs to be a quick and efficient way to locate mobile users with incoming calls. There also must be a way to ensure the user's service features go with him on the road. And all charges must show up on a single, consolidated invoice.

The answer may be to program the subscriber device to talk to a special mobility server. The mobility server looks just like another subscriber device to the stupid network. In fact, the mobility server might be software running on the home phone.

Here's how it would work: When subscribers are at home, calls are routed to their home phones. When they leave home, the wireless handset sends a message to the home phone, telling it the user is away. While traveling, the handset sends brief messages that keep the home phone informed of the user's current location. The home phone then can reroute incoming calls either by telling the originating device how to reach the user directly or by relaying the packets itself.

Clearly, significant development is required in order for Eisenberg's stupid networks to support mobility. In the end, smart devices connected to stupid networks will yield greater overall intelligence. After all, while today's average metropolitan mobile network has two or three intelligent switches, the wireless stupid network of tomorrow will contain millions of distributed processors. There will be no limit to the number of intelligent third-party servers and gateways that can be attached to that network (often without the network operator even knowing about it).

WIRELESS & IP NETWORKS According to IP telephony pioneer PhoNet, most of today's interest in combining wireless and IP telephony is in developing countries. Because they lack basic telephone infrastructure, fixed wireless links are often the only way to provision Internet and/or intranet access quickly and affordably. This can be accomplished using wireless local loop, wireless metropolitan area or satellite networks. Once fixed wireless data links are in place, adding IP telephony is a natural next step.

Tomorrow, however, things will be different. What's really holding back IP telephony is not lack of standards, but lack of performance. As private IP networks proliferate, and the Internet's backbone capacity and speed increase, more users will take advantage of IP telephony. The bottom line is that real-time IP telephony requires high-speed, low-latency, packet switching. The technology is at hand, but it has not been widely deployed yet.

Natural Microsystems already is developing the pieces that would enable wireless operators to build networks based on IP architecture. According to Chris Smith, wireless segment manager, an IP approach would provide wireless operators with three major benefits: reduced backhaul costs, the ability to employ off-the-shelf IP networking components, and faster development and deployment of new features.

In today's wireless networks, backhaul link efficiency is limited by dependence on circuit-switched connections. A cell handling 100 simultaneous calls consumes 100 voice channels over the backhaul link. But if an IP network architecture were employed, that same cell could handle 100 simultaneous calls while consuming perhaps just 25 voice channels over the backhaul link. This is because packets from different callers may be freely intermixed over the same channel.

As the Internet and intranets grow, IP networking components will become faster, smaller and less expensive. After all, IP networks consist of many pieces -- pieces taking advantage of high-volume production. INs, on the other hand, consist of a small number of large switches. The mainframes never will achieve high-volume production (although the channel cards already have done so).

The marriage between wireless and IP networks is a good one. A large percentage of wireless subscribers already use digital technology. And wireless telephony, like IP telephony, already has been slimmed down to the lowest bit rates acceptable to the broad consumer market. Plus, digital wireless technologies such as GSM, CDMA and TDMA already are pioneering integrated voice/data applications.

CDMA is perhaps best suited to IP networks. With its variable bit rates, CDMA is a natural fit for packet-switched transmission. And with its high capacity and use of channel-consuming soft handoffs, CDMA has a need for more efficient backhaul links.

The combination of wireless and IP networking will bring end users an exciting array of new capabilities. Real-time voice communications will be one simple application. Other applications will include store-and-forward voice mail, workgroup audio conferencing and audio broadcasting. Speech recognition and synthesis will permit users to communicate in the mode most appropriate to the time and place. If users need to retrieve information from a web site while driving to the office, they simply use a text-to-speech gateway. Or if they need to confer with a colleague in Japan, they might want to use the automatic language translation server.

Indeed, users will be able to exchange messages with other people without having to know whether they are carrying cellular phones, 2-way text pagers, voice pagers or notebook PCs. The subscriber devices will work out the details with the gateways.

Another benefit of combining wireless and IP networks is that users will be able to stay connected all day long. That would be impractical with circuit switching, because a circuit-switched connection consumes bandwidth whether or not it is being used. A packet-switched connection, in contrast, can be maintained using only occasional, brief messages (query: are you there? response: yes). Overlaying wireless networks on IP networks still needs work -- IP telephony in general still needs work. But the evolution is happening. The future belongs to ultra-fast packet-switched networks.

Wireless operators can jump in sooner or later, but they will jump.

Brodsky is Datacomm Research president. His e-mail address is ibrodsky@ix.netcom.com.