20 3 KEY PUBLIC NETWORK CHARACTERISTICS The variety of public network choices cited in the preceding chapter lead some potential users to despair at what seems a plethora of alternatives. Actually, specific user requirements can quickly eliminate many network candidates. This permits a more straightforward assessment of realistic alternatives. A few of those salient user characteristics are discussed here. 3.1 COVERAGE While the scale is continuous, ranging from international down to a single truck depot, the user coverage requirements discussed in this book fall into three broad categories: 1. Nationwide . 2. Regional : Examples of this might include the field sales company that works a territory from Philadelphia to Boston but has no interest whatsoever in St. Louis or San Diego. 3. Metropolitan : Taxi companies fall into this category, as do many blue collar dispatching jobs (e.g., plumbers, electricians). But many highly professional occupations such as lawyers and physicians may well be interested in only local coverage. As user coverage requirements broaden, the number of qualified service providers diminishes rapidly. Metropolitan vehicle location networks such as Teletrac are useless for statewide applications. Multistate packet networks established by regional Bell operating companies (RBOCs) and their counterparts are often early casualties The Wireless Data Handbook, Fourth Edition. James F. DeRose Copyright © 1999 John Wiley & Sons, Inc. ISBNs: 0-471-31651-2 (Hardback); 0-471-22458-8 (Electronic) when the application spans the nation. The old circuit switched cellular networks usually provide wide connection capability at the cost of a roaming fee. But the RBOCs are going digital, and no two carriers, it seems, can agree on a common standard. Digital AT&T Wireless devices will not operate on BAM networks (and vice versa); neither can work with the new PCS. Service provider candidates are classified by their network coverage capability in subsequent chapters. ARDIS and BSWD are examples of nationwide networks. But that classification does not prove required coverage. Insidious problems remain. These terrestrial systems truly cover, and seamlessly link, population centers of the United States. They do not cover vast stretches of mostly open geography in between the metropolitan areas. The executive or salesman or professional using a car to traverse Pennsylvania will be mostly out of coverage on Interstate 80, even though all the cities and most of the towns of the Keystone state are covered. Satellite carriers will cover these beautiful mountain areas but may be useless in Philadelphia. The user must first define the required coverage area, then test the candidates capability in those areas. Techniques for limiting test costs, along with many field test results, are discussed in Chapter 12. 3.2 PENETRATION Street-level, metropolitan area coverage does not ensure that contact can be held or established inside a building. Failure does not manifest itself as a knife switch; the problem becomes progressively worse the further one retreats to the interior, away from windows. Few systems are designed to specifically address this problem. One that does, ARDIS, provides a contractual performance guarantee since this is a major selling point. But excellent penetration can often be encountered in a system that has rich infrastructure in particular areas. BSWD clearly falls in this category in the Greater New York metropolitan area. Some customers have unrealistic penetration targets: Utility companies may want meter readers to transmit in deep basements, for example. This type of requirement is a recipe for failure; radio does not yet transmit through dirt. Building penetration capability can sometimes be inferred. If a cellular voice phone call can be sustained from deep within a building, then the packet data counterpart is likely to succeed, given that CDPD has been built out. If voice fails, so too will data. Some data-only carriers provide prediction tools that estimate the percentage of in-building coverage by ZIP code. These tools are usually directionally correct and can reduce much manual, tiresome labor to verify penetration. 3.3 MESSAGE LENGTH This user requirement spans the gamut from the shortest of status messages to multipage facsimiles. Very short messages, especially, may also be infrequent: a monthly odometer reading or a random door open/closed indicator. In between these 3.3 MESSAGE LENGTH 21 extremes are a range of message lengths that can absolutely force one network design to be chosen over another. Paging messages tend to be short; millions of users have long since been trained to live with only 40 characters of information. Dispatch messages are somewhat longer, accompanied as they are by location addresses, problem symptoms, or warranty identification numbers. There are a myriad solutions, including the new PCSs, which are well suited to traffic with a length less than ∼ 150 characters. Messaging traffic is longer still and can become impressively large if there are a series of message-forwarding actions that concatenate prior notes. Further, peer-to-peer messaging traffic is balanced. The outbound traffic equals the inbound, and some systems have problems with inbound activity. Conventional E-mail traffic often has long message lengths if for no other reason than the user tends to sign on intermittently, gathering several notes at a time. Both packet and circuit switched solutions abound in the messaging/E-mail area. They are very price competitive, so that the user decision is rarely based on cost alone, but on secondary factors such as coverage or the need for facsimile as well. Worldwide Web access via the Internet can produce enormous traffic, especially from the Web to the user device. There are techniques for suppressing graphic information, of course, but that is often exactly what is desired. In that case, a circuit switched solution is called for. The purest form of nontextual, graphic information is the facsimile. Data-only systems cannot handle a facsimile in the form of a pictorial representation of a piece of paper. However, it is possible to send a text message from a device to a facsimile machine. In general, with current technology, the longer the message length, the more that traffic is suited for circuit switched solutions. 3.4 MESSAGE RATES A key application characteristic is message frequencythe rate at which traffic is generated. It is sometimes crucial to make provision for the rate factor with a carrier. As an extreme example, the user interested in just a monthly odometer reading must make provision in the device to suppress registration until the moment comes to transmit. Otherwise the device may roam the nation, constantly registering on new base stations and generating considerable unwanted overhead for the carrier. Locking the device down to a specific time interval can result in greatly reduced monthly airtime charges. At the other extreme is the high-traffic generator, with 3060 messages per hour. This is usually seen in campus situations where, say, the MacDonalds order taker works the line of waiting automobiles to generate take-out hamburger requests. If the message traffic has any appreciable message length, as in MacDonalds, it may be best to abandon any hope of using a public system and install a private system instead. Every variation in between exists, and each case is different. Generally, however, high rate applications are best suited to packet switched solutions, but will be vulnerable to response time delays. 22 KEY PUBLIC NETWORK CHARACTERISTICS 3.5 DEVICE SPEED In general, sustainable bit rates fall as the user device goes into motion (the particulars are covered in Chapter 13). In circuit switched systems the modems literally slow down. In packet switched systems, retry rates rise, as the network has increasing difficulty delivering a clean message. Each network solution is designed to perform best at a specific speed. Some, like Metricoms Ricochet, can tolerate very little motion at all. It is really a desktop solution. Others, like the high-speed protocols used in ARDIS or CDPD, perform best at vehicular speeds encountered in metropolitan areas: 2040 mph. Some low-speed protocols, such as the alternative used in ARDIS, are particularly well suited to slow-speed pedestrian movement. It is rare to find a protocol that performs consistently well at speeds over 70 mph. If the application requires relatively high target speeds, it is best to clearly understand the carriers pricing plan. Some packet carriers charge only for user messages successfully transmitted; any necessary retries are absorbed by the carrier. Others charge every retry against the customers bill. Indirectly, the most obvious example of charging the user for network retries is circuit switched cellular. The user is charged for time on the system. If the messages are faltering, and modem speeds are dropping, the extended time necessary to complete the transaction shows up in the monthly phone bill. 3.6 CONNECTIVITY Most users contemplating wireless data communications approach the new networks with an installed base of application programs. To preserve precious spectrum, many packet carriers have unique protocols the new user must master before exploiting the network. Existing application programs must be converteda slow, costly task. Worse, once converted, the modifications tend to tie the user to a specific network. A flood of middleware has evolved that enables the user to convert once and gain network independence. Chapter 17 discusses these alternatives in more detail. Some network solutions require very little in the way of application conversion. Consider the laptop PC user who dials an application host over a conventional wireline modem. This user can generally connect via circuit switched cellular, and exploit wireless data, with no application change. This includes sending and receipt of true facsimiles. The packet switched solution CDPD uses the ubiquitous transmission control protocol/Internet protocol (TCP/IP). This makes application conversion a vastly simpler task but often leads to miserable performance. Once again, middleware providers have stepped in to ease the connectivity problems, essentially by eliminating TCP/IP at a gateway ahead of the wireless network. 3.6 CONNECTIVITY 23 3.7 SUMMARY Users of todays technology can make an initial sort of network services as a function of their own application requirements. Table 3-1 lists representative application characteristic and matches them to the carriers that are best suited to match that need. The chapters that follow will expand on these themes so that well informed initial judgments become the norm. Table 3-1 Application requirements versus service providers Requirement Best Match Wide-area geographic coverage Circuit switched cellular; Geostationary satellites (GEOS): AMSC, Qualcomm OmniTRACS. Nationwide metropolitan coverage ARDIS, BSWD, SkyTel2 In-building penetration ARDIS, most paging systems Short, infrequent status messages Aeris, Cellemetry Two-way paging ARDIS, BSWD, SkyTel2 Regional, metropolitan messaging traffic ARDIS, BSWD, CDPD Long file transfers Circuit switched cellular True image facsimile Circuit switched cellular Full Internet access; nonmobile situation Circuit switched cellular (price premium); Metricom Ricochet (where available) Intense message rate activity Private system Low-impact application development Circuit switched cellular; middleware for TCP/IP (most packet networks); CDPD especially for user datagram protocol (UDP) transition Voice, with modest supplementary data Nextel, PCS: Aerial, BSM, Omnipoint, PacBell 24 KEY PUBLIC NETWORK CHARACTERISTICS