Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Chapter 2 VIE PHASE IV CAPITAL REPLACEMENT PROGRAM The Air Force has decided to treat Phase IV procurement as a capital replacement program. The Committee supports that decision completely. The program properly emphasizes the much-needed modernization of the data processing equipment used to provide base level services. The Air Force's current equipment is aging, obsolescent, and more than a decade behind that used to provide comparable services to industry. Its continued use would increase the already high costs for programming, operation, and maintenance, and would not permit the expansion of capacity and capabilities needed for improved services. The committees support of the Air Force's present emphasis on hardware replacement, implying a corresponding postponement of software redesign, is an expression of realism rather than enthusiasm. The complexities and delays inherent in the procurement of data processing equipment by the U.S. government are well known. In a project the size of Phase IV, adding the complication of software change would have thrown additional obstacles in the path of the long overdue modernization effort. However, as will become clear, the full potential of the new equipment lies in its ability to provide information services in ways very different from the present software-hardware combination. Once the Air Force has adapted its current software to the new equipment, it can use the latter to provide many new services that are not now economic. An important conclusion of the committee should be recorded: because of trends in technology, the Air Force will probably be able to make future capital improvements by incremental rather than wholesale investments. The equipment proposed by either of the current Phase IV contractors will be compatible with evolutionary growth and change. By following such an evolutionary pattern, the Air Force's information services at the base level can benefit from continuing decreases in hardware costs, as have information services in industry. Thus, the economic forces that have in the past dictated high-cost, centralized approaches to computing will no longer exist. Processors with the capacity and power of the medium-sized systems now in use at Air Force bases can even now be built on just a few silicon chips. 12
13 When this revolution has finally run its course, it will be possible to create functionally distributed processing systems of great power and low cost. Because of their modular character, the Phase IV systems will probably never need to be totally replaced. The Air Force should aim to avoid another complete capital replacement, especially in view of the difficulties of the past decade {see Appendix A). ADEQUACY OF THE PROPOSED EQUIPMENT CONFIGURATIONS The family of computing equipment proposed by each of the potential contractors in Phase IV meets the Air Force's needs. Each offers substantial modularity of function and a range of options for processing, input-output control, and storage. It should be possible to select with much greater precision than before the exact amount of processing power required at a given installation. It should also be possible to expand the systems into substantial aggregations of processing and filing power capable of handling many times the presently anticipated load. Data communications interfaces appear reasonably straightforward. This is an important prerequisite to developing an effective network or system of distributed functions. Many issues involving standardization of communications protocols will, of course, have to be addressed as the system develops. It is fortunate that modern computer systems have a wide range of processing options because the Air Force may have underestimated the potential future need for computer services. The forecasts shown to the committee seem to project a computing load growing roughly linearly with time. Industry, however, teas experienced more nearly exponential growth, though without the artificial constraints characteristic of the Air Force's older base level systems. When Phase IV's more flexible, capable, and economical systems are introduced, new applications will become not only possible but also economically justified through superior service to users. Therefore, the committee believes that demand will be substantially heavier than forecast by straight-line methods. Phase IV will share resources wherever justified to cut costs. Some bases will use computing services provided by data processing installations located at regional centers, thus reducing duplication of equipment, facilities, and personnel. These savings will be offset by added communications costs. Whether regiona~ization can reduce overall costs will depend on the specific costs of people, computers, and communications. However, the economics that have justified regionalization in the past are becoming less valid with time. In earlier generation computers, the high costs of central processors and high-speed memories dictated centralizing of computing resources. With new technology, processors and high-speed memories cost much less. This allows processing to be performed where data files are logically maintained. Lower cost of higb-speed memories allows more sophisticated operating systems to be used, even by smaller processors. Software systems that rely less on batch processing techniques will
14 reduce manpower requirements. The Air Force should analyze costs and their trends to determine how much regionalization to adopt under Phase IV. THE PEASE IV PROGRAM BEYOND TRANS ITION The Phase IV program management should be able to deliver operational sets of the transition software on both competing systems. It should also be able to convert the remaining software expeditiously, using a variety of automated tools. When the transition is complete, the Air Force will have acquired contemporary equipment. However, except for some software modernization achieved by the Phase IV vendors, the Air Force will be using programs originally designed for outdated hardware. Longer term use of these old programs will prevent the new equipment from being operated as efficiently as possible, as well as-probably requiring more manpower. For example, in the desire to minimize software changes in the transition systems, operations that still depend on punched cards and magnetic tape or other sequential file structures may be retained even though on-line access and data base management systems would offer better service at lower costs. The transition systems will probably continue to rely significantly on batch processing. Thus, they may continue to require more manpower than would be needed if the applications were reprogrammed to serve users interactively directly from remote terminals. The Phase IV system's technological capabilities should not be the reason, however, for redesigning all the software. Many existing applications, which produce periodic reports, should probably be kept. Unless it would either lower costs (e.g., by reducing paper handling or personnel) or improve services to users (e.g., by giving supply and maintenance personnel access to common computer files), software redesign should not be undertaken. Programmers are a costly resource in short supply. Phase IV software will be covered later in this report. It is important to recognize, however, that the Phase IV systems, once installed, can and will provide the basis for greater efficiency and management responsiveness. The new systems can connect to and communicate with other information systems, both on the base and at other locations, with low additional investments in equipment. The new Phase IV computers will lend themselves to distribution of computing functions and local area networking--both important evolutionary steps to prevent the need for another capital replacement. The Promise and Challenge of New Technology Information processing is changing more rapidly than almost any other technical endeavor. Because of continual innovations in hardware and software design, computer systems put into service as recently as five years ago may be functionally obsolete today. For example, the
15 hand calculator--a radical innovation in the early 1970's--could perform many computer functions for less than $300. It made its mechanical predecessors obsolete. Today's hand calculator of equivalent power costs less than $30, most of that for the display and keyboard. Next year, the same $30 might buy a combined digital watch and programmable calculator. Thus, technological trends need to be considered in planning for future information systems. Manufacturing processes using very large scale integration (VLSI) techniques can produce the same logic on three small silicon chips that now exists in a medium-sized ~mainframe" computer. The $1,000 "mainframe" computer is now in sight and will probably shrink to a $100 mainframe later in the 1980s. VLSI techniques have also radically decreased the costs of main computer memories. In earlier generation systems, the high cost of memory tended to limit the number of applications or of simultaneous users a processor could handle. Now, even inexpensive computer systems can implement sophisticated operating software for multi-user applications. Though the mainframe costs alone do not define total system costs, this drastic rewriting of the cost/performance rules will revolutionize information systems. The current practice of centralizing computing hardware in one room can hardly be justified when the same power can be distributed in quantity at very low cost to locations where the processing is needed (e.g., the finance office, maintenance shop, flight line, or supply room). It may be difficult to understand the relevance of the "number of computers at a base" when they are counted in dozens and priced in the hundreds of dollars each. When processors can be obtained that have the cost and size of today's hand calculators, but the power of a current mainframe, then functions can be added to a system by procuring additional specialized subsystems each having low cost and high capability. Planning for systems of this type is underway in the business community. Docal area data networks (e.g., Ethernet, Wang Net, or Hyperchannel) promise effective and economical integration of dissimilar specialized processors into a flexible aggregate. Moreover, this revolution is not confined to hardware. Software systems will evolve to include functions such as data management and retrieval as parts of their own processors. They will also perform these functions for other processors communicating with the same data bases. Finally, the processors need not all be made by the same manufacturer or be located in the same room. Thus, they will have a potential for evolutionary growth toward highly flexible base level systems whose components can be tailored to service the needs of a given base. The Air Force must recognize these trends as inevitable and desirable, and should meet them head on. Without a coherent plan, information processing at the base level might drift into an uncontrolled and uncoordinated proliferation of independent m~niprocessors and microprocessors in the personal computer class. Because of their low cost, these appear highly effective for satisfying user needs, but they must be regarded as part of an
16 integrated base-level system and added within the framework of a coherent plan. For example, word processing systems containing shared logic and terminals will undoubtedly acquire more powerful computational capability. Once capital replacement is complete, Phase IV computers could serve as hosts for integrated base level systems. The emerging technology of distributed processing, coupled with bigh-performance local data networks, could connect functionally specialized systems at low cost into cooperating base networks. If this opportunity is missed, Phase IV systems could become splendid relics unable to adapt to changing times. Phase IV should be regarded as a vehicle to maintain the information systems at the leading edge of technology and to offer greatly improved and better integrated services to base level management. The need for closer coupling between data automation and communications deserves special attention. In a distributed system, communications is part of the system rather than a necessary but separate function, as it is now. It will be difficult to connect separate processing elements into a local, base level distribution network unless the data automation and communications communities jointly define the system characteristics. There will, of course, continue to be a need for centralized host computers; these will continue to communicate with higher echelons or with remote centers, such as the Air Force's Manpower and Personnel Center or the Accounting and Finance Center. Intra-base networks can be created to distribute data by combining conventional and advanced techniques. Today's point-to-point wiring may give way to more flexible switched connections using digital private automatic branch exchanges (PABX). Using coaxial or fiber optic cables, processors will be connected with terminals on a common data bus at high speed and relatively low cost. The Air Force may have to consider new organizational approaches to defining composite requirements for local systems that integrate host processing (Phase IV equipment), word processors and electronic mail services, and a pervasive local data network. At present, these are the domains of three relatively independent communities, but the interests of these communities will tend to merge as an outgrowth of the new technology. Personnel Without exception, both the system users and the data automation community within the Air Force have many dedicated and knowledgeable professionals. They are producing workable software systems despite cumbersome batch operations, too few terminals, and the absence of modern software development tools. Users in all functional areas seem well aware of the limitations of the present systems, but have adapted to them and are enthusiastic at the prospects of a modernized system.
17 However, contemporary equipment and improved capability will highlight the Air Force's critical problem of too few highly trained personnel. This gap would prevent the Air Force from realizing the full potential of the Phase IV system. For their definition, development, and effective use, new services and applications will require a cadre of software specialists, fully trained in modern information systems. The Air Force should examine its training programs, reenlistment incentives, and recruiting practices to improve the retention rates for data automation specialists at all levels. The emphasis should shift from filling slots to building quality and experience. Obtaining high-quality graduates, particularly in computer science, should receive more emphasis in officer recruiting. Crossflow from other specialties into data automation should be viewed as less desirable than retention of experienced enlisted personnel. In summary, because software rather than hardware will determine the success with which Phase TV and its successor systems provide efficient, versatile services to users, and since quality software can be produced only by highly trained and experienced specialists, it is now more important than ever that standards of quality and training be carefully examined and improved by all possible means. In addition to ensuring continued availability of highly qualified computer professionals, the Air Force also needs to plan for a gradual change in how these people support base level automation. At present, data automatons assume total responsibility for meeting the computing needs of the users located on an air base. Users are regarded as consumers, not creators, of services. The evolutionary change that must be planned for will come about when users begin to acquire more independence in their own computing affairs. If the Air Force adopts an approach involving more autonomous, functionally distributed processing at base level, and if users can do some of their low-level programming on distributed small computers, data automatons can begin acting as advisors and disseminators of programming techniques and useful software to help the users get the most out of their new autonomy. Software: The Key to Improved Services Current software systems that operate at base-level installations represent a substantial investment of human resources. After this software is adapted to the new equipment, many additional systems will be developed to meet expanded user needs. The ability to implement new services will depend on the productivity of the data automation specialists. Like its counterparts in industry, the Air Force should emphasize programming productivity as the key to the successful creation of new and modern software systems capable of exploiting the new equipment. Because hardware costs will diminish over the life cycle of the Phase IV program, attention will increasingly focus on programming productivity and the maintainability of software systems. Personnel costs, not hardware costs, dominate today and will probably continue to do so.
18 Several aspects of software technology should be fully studied and exploited to contain the cost of developing and maintaining new systems. Some permit programmers to be more productive, some lower maintenance costs, and still others reduce software costs by enhancing interchange and sharing of software among commands. Base-level software systems should be produced with the aid of specialized software development facilities at the Air Force Data Systems Design Center and at major commands. These facilities should be linked in a network to permit software sharing. Also, structured programming techniques should be aggressively adhered to and data base software systems should be used for efficient file management. Finally, query and report generation languages should be improved to increase their use and thus reduce or eliminate the need for programmers in many of the simpler, user-specific applications. Maintaining a Focus on Evolutionary Growth Even without the problems imposed by the cumbersome and time consuming Federal procurement process, the Air Force has many reasons to avoid major capital replacement programs. Replacing large numbers of common computer systems worldwide can be prohibitively expensive, time consuming, and disruptive. Still, obsolete equipment no longer in production needs to be completely replaced. Because this equipment supports many data processing programs for different users at many locations, all are affected by the replacement. With the new generation of computing equipment now introduced, however, it will be possible to avoid such being expensive traps. The key will be found in the drastic reduction of costs of processors, memories, and other electronic components. Entire collections of programs and services no longer need to be combined in the same central processor. Information processing is moving toward distribution of functions into separate, small pieces each performing a function for which it has been designed. For example, communications are being handled now by specialized front-end processors. If communications networks should significantly improve, front-end processors might be exchanged for more capable models without affecting central processors or application programs. A wholesale capital improvement program might not be necessary to improve communications. Similarly, recently introduced Back-end processors" or Data management engines" (specialized processors) can retrieve and manipulate data. They Can thus relieve mainframe processors of that duty. Improved data structures or retrieval speed can oe orougnc about by exchanging back-end processors, again without signif icantly altering other elements of the system. Another opportunity for evolutionary growth will arise from a different kind of distributed processing. In addition to providing distributed computing processors (e.g., communication or data base), specialized processing can be distributed by user functions. For example, a specialized base level subsystem could be designed to i
19 support civil engineers. Such a system could be located close to and configured for their software. Because the subsystem would have local communications, it could be considered part of the base-level system. Such systems need not be expensive: a civil engineering system might cost $50,000, whereas a system to support the base pharmacy might be only $20,000 (hypothetical but possible figures). Neither of these hypothetical systems would require manpower per se. Each would be operated entirely by those who use its services on their jobs. The important issues are the overall cost of the system, including software, and its ability to support the users adequately. Thus, the days of monolithic systems are, fortunately, numbered. Compatible processors from the Phase IV family will range from very small to very large. Processors from other vendors may be added to a base-level local network to support functions while communicating with others in the network, including the central host systems. The advantage to the Air Force is that this concept will permit orderly, nontraumatic growth and change; evolution will be possible by adding or changing functional processors in a less disruptive way. Control over the base-level networks is a concern for managers. Presently, data automation and communications organizations are separate entities at each base. If distributed processing networks are introduced to replace the current centralized systems, data communication becomes part of the system rather than merely a service function. The data automation community will need to take the initiative for directing the orderly adoption of such network-based systems, while permitting users the flexibility that the system promises. Centralization of all applications on a big mainframe in a single computer room, once prized as offering economies of scale, should be regarded as contrary to evolutionary growth and not cost-effective. It should be avoided in favor of distributed functions, local communications, and dissimilar but compatible small processing units.
