A Personal Trek into the World of Wireless Data
[Note: I recently found this article that I had written 18 years ago in a publications database on the web. For context, this article was written 12 years after the first introduction of the IBM PC, and two years before the initial public availability of the World Wide Web. Some of the topics of discussion are quaint today, such as the focus on mainframe terminal-based applications and on the proprietary wireless networks that prevailed at the time. But the fundamental observations I made then are still true today. Wireless data applications have been a long time coming. – WS]
Wireless datacom is a jump-shift from conventional datacom. A look at some of the fundamental differences.
As technologies change the face of business communications, those who develop, market, manage and use new communications technologies also must go through a rite of passage to understand, interpret and implement new solutions. One particularly challenging transition is the shift from the world of conventional, wireline data communications to the wireless data communications world.
Two factors are stimulating interest in wireless data communications. One is the explosive growth of cellular service. Technology brought a unique and much-needed solution to an area--mobile voice communications--that had been poorly served in the past.
The second factor is the growing disparity between the automation tools available to those who primarily operate from fixed locations (i.e., traditional office desktops) and those who are "mobile." This disparity was not an issue when corporate information systems were primarily host-based transaction processing systems, but personal computers and workstations have altered expectations for employee productivity and effectiveness. As a result, organizations are now investigating ways of bringing mobile employees into the information network.
Last year, I made a personal and professional transition into the emerging world of wireless data communications. Even though I had worked extensively with large corporate data communications environments--IBM SNA, NetBIOS and Netware PC local area networks and "open systems" communications (i.e., TCP/IP and X.25)--my years of experience with traditional data communications proved to be of limited value.
Many fundamental differences exist between the new wireless data communications world and the familiar data communications environment from which I had come. I will focus on wireless data communications in wide area or metropolitian area data networks, as opposed to wireless LANs.
Scarce Spectrum
One of the most striking differences between traditional datacom and wireless data is that the basic, fundamental medium is extremely limited. Only so much radio spectrum is available, and unlike copper or optical fiber, there isn't any more being made. This fact affects all aspects of wireless communications, including its availability, cost and data rates.
The total amount of raw usable spectrum bandwidth available for wireless data communications is roughly equal to 1,700 to 2,000 T1 circuits, 250 to 300 Ethernets or 25 to 30 FDDI networks. While the wireless datacom industry is developing techniques for more efficient frequency reuse and allocation, data compression and high-efficiency protocols, the laws of physics will prevent us from converting to a totally wireless communications world.
Some industry pundits argue that spectrum isn't really scarce but it used inefficiently. While it is true that spectrum in the U.S. isn't used efficiently, the reality in today's market is that virtually all of the most usable frequency bands have already been allocated, either to a specific user or to an industry.
Moreover, the mechanism for allocating spectrum in this country is intensely political, and the rules of the game have been set up such that it is a win-lose situation: If you want access to spectrum, you have to dislodge whoever is currently using it. This process is expensive and uncertain.
In many cases, the incumbent owners of spectrum have long-standing relationships with the regulators, and regulatory authorities have a long history of favoring voice over data communications. The bottom line is that spectrum for data communications will remain scarce.
The scarcity of spectrum is part of what makes wireless communications so much more expensive than traditional landline communications (see Figure 1). Corporations should be wary of implementing wireless data communications casually or in volume, because it can quickly add up to serious money.
Telecommunications managers who pay the bills for cellular telephone service can appreciate how quickly wireless communications expenses can run up. In addition, many wireless data networks charge by the packet, which can accelerate costs even more.
The Federal Communications Commission's historical philosophy toward spectrum allocation is not consistent with efficient wireless data communications services. For example, spectrum tends to be allocated in channels that represent the spectrum capacity needed to carry human speech--about 25 kHz. The present practical maximum amount of data that can be pushed down such a channel is 19.2 kbps, hardly significant by conventional telecommunications standards.
Even though wireless data seldom achieves this maximum because of these networks' high errors rates, the effective data rate is usually enough to support the kinds of interactive terminal-oriented applications that are in common use today. Moving larger amounts of data, such as file transfers and faxes, tends to happen at speeds characteristic of dial-up telecommunications rather than the rates supported by LANs or backbone landlines. If the objective was to extend existing terminal-oriented applications out to the field using wireless communications, the scarcity of spectrum would prevent enough channels from being available to support a significant population of terminals going on line concurrently.
Another problem is that a single slice of spectrum is usually allocated to several geographically separated users. Various users in Seattle, Dallas and Boston may receive an allocation for the same frequency because their transmissions could not interfere. Most of these frequencies were originally allocated to local-only services, such as public safety and taxi dispatch, and little attempt was made to make coherent and consistent frequency assignments across the nation.
As a result, it is difficult to get a single uniform national channel assignment. Thus, a company may have data radio equipment in one city that won't operate in another city. While today's data radio equipment is usually "frequency agile," and thus able to operate on a broad range of frequencies, this capability doesn't come free--it increases the cost of the already expensive equipment. Moving data over the cellular telephone system and over certain national packet data services, however, provides national service on the same frequency bands.
Hostile Environment
Reliability problems are another surprise waiting to happen. Wireless data is not as reliable or accessible as data moving over wireline telecommunications or LANs. The radio environment has significant signal quality problems compared with copper wire or fiber optic media, and these problems are particularly troublesome for data communications, where small interruptions in the link can corrupt data or even drop the link.
Radio signals attenuate sharply over distance, and signal strength typically falls proportionately to the distance between the sender and receiver. In wireless communications, the sender and/or receiver is often moving, and it is impossible to install signal repeaters or amplifiers between them. As a result, a rapid decrease in signal strength is unavoidable.
This significantly limits the practical distance between sender and receiver. The equipment to increase that range--to detect a weak signal and to amplify and process it to a usable form--becomes more expensive as the desired separation between sender and receiver increases.
The radio environment is much noisier than the telecommunications environment, and a variety of factors make it difficult to distinguish the sender's attenuated radio signal from other radiated spectral energy in the band being used. Some of this noise is natural (like the interference from sun spots), while some comes from artificial sources--e.g., licensed in-band transmitters that are not operating properly and are spilling energy into the signal band.
Interference can also come from the sender's own signal as multiple reflected versions arrive at the receiver's antenna at different times (i.e., multipath effects). The strength and nature of interference is changing constantly, and when it is strong enough to disrupt the ability to distinguish signal from noise, bits of the data stream can become lost. This is especially common when one end of the link is moving. Again, more capable and expensive equipment can overcome some of these problems, but there are limits to what the technology can achieve.
The effects of range and interference, combined with the ever-present limits on available bandwidth, severely constrain the data rates that wireless systems can sustain. Wireless local area networks can support data rates of millions of bits per second, but the rates are still far below those provided by high-speed LANs like FDDI.
Conventional data protocols are not practical in the wireless environment
The difference in data rates is even more pronounced for metropolitan and wide area networks. In a world where wide area telecommunications networks often are designed with T1 as the minimum unit of capacity and new low-cost modems are achieving 56 kbps over dial-up connections, wireless data networks are struggling to deliver 19.2-kbps links to mobile users.
Unfortunately, the practical throughput on wireless networks is usually about half of this "rated" capacity because of the high overhead of signaling and error correction protocols. This will significantly influence the architecture of potential wireless applications. Many companies will find that it is neither feasible nor affordable to extend even terminal-oriented interactive transaction networks out over the wireless environment unchanged, let alone extend native connections to LANs.
As if the wireless data environment did not offer enough challenges for users, it is also not dependable. Those who "rent" airtime--which you will do unless you are one of the elite group that already owns spectrum--cannot assume that a wireless link will be available on demand. Most cellular telephone customers who live in large cities and use their phones during rush hour are already familiar with this phenomenon.
Spectrum is indeed scarce, and there are more people who want to use it than the available bandwidth can support. Even the largest wireless packet data network in this country is reported to hit capacity limits during peak use periods. The lack of dependable availability is likely to worsen as wireless data services become more accessible.
Yet More Data Protocols
Another surprising aspect of wireless data networks is that they use protocols that are entirely different from anything else you have ever seen. Conventional data protocols are not practical in the wireless environment because the bandwidth is too scarce and because normal bit error rates are too severe.
To maximize the bandwidth for user data, wireless protocols cannot afford to use large packet sizes, large packet or frame headers, frequent positive acknowledgments, inefficient error recovery methods (i.e., send the packet again) or other protocol overhead such as address and route broadcasts, or excessive link and session synchronization. Conventional data link and transport protocols--such as SNA, TCP/IP, NetWare, Ethernet and X.25--won't be used because they are spectrally inefficient and inadequate to deal with the error-prone radio environment.
Even modem-based data communications over standard cellular telephone channels increasingly use custom wireless protocols such as MNP-10 and modified versions of V.42bis. While wireless packet data network protocols are unique, most of these networks provide somewhat transparent gateways for the corporate network using conventional protocols such as SNA/SDLC, X.25 or bisync. The connection at the mobile radio modem is somewhat different, however, and often requires the end user device to incorporate some awareness of the wireless network protocol.
Wireless data protocols are newer than their wireline counterparts, and thus, not surprisingly, they tend to be less mature, robust, functional or integrated than their older siblings. For example, network managment functions are generally not supported into the wireless network, let alone across it.
With the exception of the cellular modem protocols, most wireless data protocols are proprietary to the network or equipment vendors. This contrasts with the trend toward "open systems" protocols that are controlled by vendor-independent standards organizations. This level of maturity does not yet exist in the wireless data communications industry. The consortium behind the evolving protocol for the Cellular Digital Packet Data network is dominated by network service providers.
While most of these proprietary protocols have published interfaces, they are still vendor-controlled, and in this regard they are no more "open" than SNA or NetWare. Most users will find the wireless data environment less "open" and more enslaved to the proprietary interests of the vendors than their corporate information networks.
The consequence is that users who are entering the wireless data world will have to learn about a different communications environment with new architectures, capabilities and protocols. While many of the concepts and techniques of classic datacom also will apply to wireless, the transition will be confusing, frustrating and challenging for a while.
The Need for Client/Server
Because wireless data communications is expensive and prone to errors, it is imperative to send the fewest bits possible through the air to perform the application. Looking at it from this context, most companies will conclude that they will not be able to run existing applications just by moving the operator out to the end of a wireless link.
Mainstream corporate transaction applications usually work interactively with operators using terminals. These applications send full screens of data as character streams, and often include additional bytes for screen formatting and other user interface functions. Many interactive terminal applications echo characters back to the screen from the application or resend the entire screen contents just to update one field.
LAN-based applications tend to move even larger volumes of data between the file or application server and the user's workstation. However, the cost of operating applications in a manner that sends a lot of data over the wireless connection quickly becomes prohibitive.
While the use of compression techniques can help reduce these costs, they are only a partial answer. Most companies will need to rewrite existing applications to migrate to a wireless environment. To minimize the data sent over the radio link, the application should send flags, codes and preformatted data rather than English-language character streams.
Applications should send dynamic data to a mobile user workstation that can interpret the coded data and manage the user interface locally with predefined screens. This implies an intelligent workstation for the mobile user, which further implies closely designed cooperation between the central application and the user's interface workstation. If this sounds like a client/server architecture, it is.
The original intent of the client/server architecture was to take advantage of the processing power of increasingly low-cost user workstations to handle local, user-specific processing tasks--e.g., graphical user interfaces, data formatting and editing, and local data manipulation. Under this concept, central application servers can be dedicated to functions that need to be centralized, such as core application and data-base transaction processing.
Adopting this architecture of application-to-application communication between the central server and the user workstation tends to reduce the volume of information that needs to flow between those two systems. A properly designed application can even more effectively minimize the data movement between the client and the server.
Most companies will need to rewrite applications to migrate to wireless
The client/server architecture is ideal for developing spectrally efficient, cost-effective applications for the wireless data environment. While corporations have been conservative and deliberate about adopting the client/server architecture in their existing information systems, the benefits of client/server are much more compelling in the wireless environment.
Looking to the Future
Wireless data applications will benefit many organizations as the need grows to provide large numbers of mobile workers the same automation advantages as office workers. The strong growth of portable computing will only amplify the need to tie mobile workers into the organization's information system.
This new wireless data communications environment presents unique challenges and pitfalls. Because spectrum is scarce, wireless data will never be cheap or easy to use. It will always require that the organization's systems and communications architects design applications and networks specifically for the radio environment.
Implementing applications in the wireless data world is difficult, but things will get better. While most of the companies that serve the wireless data market come from the voice side of the industry, they are getting increasingly sophisticated about computer systems and data communications. The FCC has developed a progressive, open-minded stance toward wireless communication, and it has actively fostered technological innovation and responded to market needs. These initiatives will stimulate revolutionary changes in corporate information systems similar to those that came from the introduction of the personal computer and the local area network.
