FAS Homepage | CyberStrategy Project | Search | Join FAS


Cyber Strategy Project

CyberStrategy Background

This Background Paper surveys current trends and development in information systems generally. It is a supporting document for our CyberStrategy Implementation Plan, which is intended to provide our community with the resources to implement our Cyberstrategy.

Contents

  • A - Stations

  • B - Networks

    1. Local Area Networks - LAN
    2. Wide Area Networks - WAN
  • FIGURES
    1. Desktop Evolution
    2. Annual Computer Sales
    3. Installed Base of Computers
    4. Household Market Penetration
    5. Range of Estimates of Market Penetration
    6. Home & Office Sales
    7. Home & Office Installed Base
  • TABLES
    1. Station Evolution
    2. Station Pricing
    3. Network Evolution


  • New and emerging desktop computing implementations are providing small non-profit organizations with computational capabilities previously confined to government agencies and large corporations. This democratization of computer power will substantially enhance the ability of the public interest sector to influence the national policy process.

    These powerful desktop implementations are rapidly spreading into the homes of millions of American citizens. The convergence of office and home computers offers new and exciting opportunities for membership participation in the work of non-profit organizations, and greatly enhances the potential of individual citizens to influence the national policy process.

    These opportunities require continuous improvement in information systems by the advocacy community, maintaining parity with corporate and government capabilities, and our adversaries in the policy arena. Continuous improvement is also required to engage the energies and match the achievements of the emerging cadre of independent cyber-activists.

    The past decade has witnessed dramatic evolution in information systems. Over this period, new generations of computer platforms have been introduced at intervals of two to three years, and new software applications at intervals of two years or less. At any given time, a new state-of-the-art computer has typically cost in the neighborhood of $3,000 to $5,000 [excluding software and installation], though its resale value falls by about half each year after purchase.

    The tables at the end of this document portray [in general terms] the past and future progress of work-station and network information systems, with each "tier" representing roughly a two year intervals of current state-of-the-art. Each matrix includes rows that list implementation tiers, labeled for convenient reference as various metals, and columns that describe the particular hardware or software required to reach that level. The metals metaphor provides a neutral nomenclature that covers both PC and Mac computers -- too frequently computer generations are discussed simply from a PC perspective, which overlooks the significant Mac legacy in our community. In each case, the dates are for the point in time at which these implementations would represent a cost-effective acquisition for our community. Inevitably, the initial purchasers of new computers pay a substantial premium of several thousand dollars for "being the first kid on the block" to buy a new model computer, and prices typically fall to $3,000 to $5,000 within a few months after initial release to the market. The "Electrum" implementation is currently about the best cost-effective implementation on the market today, while the "Tin" implementation was the best-buy a decade ago.


    A - Stations

    Desk-top personal computers have revolutionized the work of non-profit advocacy organizations. Contemporary personal computers have powerful capabilities which were previously restricted to large corporations and government agencies. Today, top-of-the-line desktop computers rival the capabilities of the most powerful supercomputers of two decades ago.

    [Evolution Graph] The radical improvements in personal computing have been obscured by the substantial continuity in the external appearance of the desk-top work-station. From the outside, today's station looks much like that of its predecessors -- a keyboard, a box about the size of a small suitcase, and a monitor, all in any color you want as long as it is beige. But these external similarities have masked profound quantitative and qualitative internal improvements. Although no single performance benchmark fully captures these developments, the speed and performance of today's desktop computers are anywhere from 100 to 100,000 times that of personal computers of fifteen years ago.

    The development of desk-top work-stations has proceeded through at least three major phases over the past fifteen years.

    The division between computer "boxes" themselves and peripherals/application software that can be used by that box serves as a separator between some of the tiers. The boxes and the associated level of central processing unit (CPU) cost the most, but the less expensive peripherals and applications greatly expand the capabilities of the basic box. Only with the peripherals and applications is the full potential of the particular CPU exploited. The implementation matrices include tiers based on CPU box level (elemental metals in the table) and, between some of those tiers, intermediate tiers (alloys of the elemental metals) based on acquisition of the relevant peripherals and applications [along with improved versions of the basic CPU platform].

    The growth in performance of these newest computers is exceeded only by the growth in hardware and software and software requirements.

    There is no escape from the reality that older computers [older than "Silver"] are simply incapable of supporting these new information systems and applications.

    [sales graph] During the 1980s, annual sales of desktop computers for office use grew at a rate of about 10% each year, from five million units to ten million units. And in the early 1990s, annual sales increased dramatically, with home computer sales growing at 20% annually. Total sales surpassed twenty million units in the mid-90s, and are expected to approach forty million units by the end of the decade [according to a study by the consulting firm Dataquest], most of which will be platforms more sophisticated than today's Pentium [according to projections by BIS Strategic Decisions]. Our analysis is a composite of the projections provided by these two firms.

    [sales graph]The current boom in computers sales derives from a convergence of office and home computing capabilities. A decade ago, the computer market was fundamentally divided between office computers costing anywhere from $5,000 to $8,000, and much less capable home computers typically costing between $1,000 and $2,000 [in 1995 equivalent dollars, adjusted for inflation]. While desk-top computers brought powerful new capabilities to the office, home users were relegated to playing simple games, typing short letters, and storing recipes. Today, both home and office users have converged on highly-capable systems typically costing about $3,000. The profound expansion in the range of capabilities of home computers accounts for the greatly expanded ownership of these systems, as both home and office systems provide equivalent applications.

    Mac implementations have been consistently priced higher than PC's with equivalent performance. This reflects the greater "user-friendliness" of the Apple computers, which are ready to run out of the box, in contrast to PC's, which typically require non-trivial tuning and adjustment before becoming fully operable. This set-up cost, paid either in user or consultant time, results in the two implementations having roughly equivalent costs, when all is said and done.

    This price difference highlights the importance of the computer's Total Cost of Ownership, which includes the post-purchase costs of training, maintenance and upgrades. According to a study performed by The Gartner Group, the total cost of ownership of a new computer in 1995 is approximately $40,000 over a five-year lifetime, double the $20,000 lifetime cost of a computer in 1987. This study assumed higher pay-rates than typically prevail in the non-profit community, but even after adjusting for this factor, the initial capital outlay of $5,000 to purchase the computer is only a fraction of the total cost. Thus the apparent savings from purchasing "sunset" systems [such as the 486 or Quadra systems in 1995] are negligible compared to total cost of ownership.

    The Electrum Pentium/PowerPC implementation is certainly not the last word in computers. It is already foreseeable that new implementations will reach the market every couple of years for the indefinite future. Capabilities which currently cost as much as a new car will come within reach of non-profit organizations and individuals by the end of this decade:

    The next-generation Gold systems will offer performance equivalent to the Cray-1 super computer of two decades ago. And desk-top systems on the far horizon, such as the notional Iridium will provide performance equivalent to the Cray Y-MP supercomputer of the late 1980s.

    [computers] Installed Base of Computers In 1995 only a small fraction of installed desktop computers were fully capable of supporting sophisticated 32-bit multi-tasking applications, which require Pentium/PowerPC platforms. However, as a result of greatly increased annual sales levels, by the end of this decade most installed computers will support such applications, as well as more advanced implementations which have yet to be identified. These projections are based on our composite analysis of forecast sales, and are consistent with other estimates of the existing intalled base of computers.

    [computers] Most of these computers will reside in the homes of individual citizens. By the end of this decade, the aggregate computational power of home computers will rival that of all desk-top office computers in public, private, and non-profit organizations combined. In the 1980s, typically the staff of a national organization had computers, and the membership did not. By the end of the 1990s, many organizations will find that the aggregate computational power of the national staff will be small relative to that of their membership.


    B - Networks

    The development of individual workstations has been matched by the development of networks of computers. And existing computers from before the Bronze-age have limited, and increasingly no, capability to exploit these advantages of computer networks. Local Area Networks [LANs] links stations and staff within an organization, while Wide Area Networks [WANs] such as the Internet link stations, staff and others at different organizations and locations.

    Computer networks provide a big boost in productivity for organizations. They enable organizations to share resources, such as printers, modems, and drives within organizations, and enhance coordination within and among organizations through file-sharing, E-mail and group scheduling. The most powerful feature of networking for advocates and analysts is that resources and information from around the organization (through local area networks or LANs) or around the world (through the mother of all networks, the Internet, and other wide area networks or WANs) are directly available through the person's desktop PC. The desktop paradigm of networking recognizes that for organizations and individuals to truly leverage the benefits of computers and digital communication, functions and capabilities should be available directly from the individual's PC. Only in this way can these technologies become integrated into daily work and become as routine as the phone on your desk. As a result of these powerful capabilities, computer networks have become common-place in most enterprises.

    1 - Local Area Networks - LAN

    A Local Area Network [LAN] of computers enables an organization to combine the skills of different people and the power of different equipment, regardless of the physical locations of the people or the equipment, and by doing so to reap enormous benefits. A well-designed LAN enables staff to instantaneously and effortlessly collaborate, view, change and exchange information. With a computer network they can share the same electronic files from their own computer without copying or transferring files. With LAN integrated applications, such as scheduling and task management, network users can collaborate with unprecedented ease.

    A Peer-to-peer Local Area Network enables a small number of networked computers to function as both a server and a workstation. In a peer-to-peer network the operating system is installed on every networked computer; this enables any networked computer to provide resources and services to all other networked computers. For example, each networked computer can allow other computers to access its files and use connected printers while it is in use as a workstation.

    A larger number of computers can be linked into a Local Area Network using a client-server arrangement, of which Novell NetWare is the most common implementation.

    Ethernet is an inexpensive LAN infrastructure which supports both peer-to-peer and client-server architectures. Ethernet provides exchange of data at rates of 10 megabits/second [Mbps], with adapter cards costing approximately $150 per workstation. The Ethernet 10BASE-T system is designed to used voice-grade telephone cable that may already be installed. Many sites choose to install higher quality "Category 5" cables and connectors to support 10BASE-T, and which provide the infrastructure for the newer 100-Mbps Ethernet media systems, facilitating future increases in network traffic.

    Although our community started networking in earnest in the last few years, many organizations remain unconnected, and many others currently have obsolescent systems. The problems facing organizations wanting to network are twofold; first, organizations are not sure which networking option will meet their needs and, second, networks can sometimes be daunting to install and administer for non-technical staff.

    The market for client-server network operating systems is currently dominated by Novell, which runs over 75% of server networks. Novell is available in two main flavors - the 3.x generation (Novell 3.11 and venerable 3.12) and 4.x generation (Novell 4.11 being the latest). Both types integrate well with Microsoft Windows and provide reliable performance. Windows NT is making inroads with its superior performance, easy maintenance and aggressive pricing.

    Case Study: Council for a Livable World

    Novell 3.12 and Internet router setup provides an example of modern server networking and desktop integration.

    Pros
    User management, data backups and network troubleshooting are centralized, making the server model often easier to administer than its peer-to-peer equivalent Printing to a remote printer and file transfers are generally faster and more reliable. New capabilities added to the server are available to all those connected to it.
    Cons
    Client-server setups are generally more expensive than their peer-to-peer equivalents. A reliable computer must be dedicated as a fileserver. Although a good value, these systems are still more expensive than peer-to-peer networks (especially for smaller networks of five users or less where the "price per seat" of Novell or NT is quite high). The operating software usually requires a computer specialist to install and configure. A failure in the fileserver can bring a whole organization to a standstill.

    Note that new types of network engineering are increasingly combining the best of server-based operations (performance, security) with the strengths of peer-to-peer networks (flexibility and cost) by running both server-based and peer network operating systems simultaneously. These hybrid networks could well be the future of client-server networking.

    A peer-to-peer network links computers together to share data and peripherals. Each computer can act as a information repository or retriever; its hard drive or peripherals are available to others on the network. Peer-to-peer networking is dominated by such players as Microsoft (with Windows for Workgroups 3.11 and Windows 95), LANtastic (the 5.x and 6.x generations), Banyan Vines and the AppleTalk. All are priced attractively especially for smaller offices or workgroups.

    Case study: The Federation of American Scientists

    Windows for Workgroups network is a demonstration of the power and cost-effectiveness of peer-to-peer networking.

    Pros
    Peer-to-peer networking is cheap. Packages such as Microsoft Windows for Workgroups or Windows 95, which come bundled with many new systems, already have networking capabilities built in. It is particularly cost-effective for small organizations that may not need the capabilities of a dedicated fileserver. Newer peer-to-peer networks, such as LANtastic 6.x or Windows 95, also provide better management tools and easy to use interfaces than their older incarnations.
    Cons
    Peer-to-peer networks tend to be harder to troubleshoot and are less reliable than client-server setups. Reliability problems are especially acute for daisy-chained installations where workstations are linked together in a chain. A failure in any one of the workstations incapacitates the whole network. Newer installations using hubs (where computers are separately linked to a central console) have helped eased reliability concerns. Peer-to-peer network performance is still sluggish when transferring larger files (such as sharing a database) and tend to require more powerful workstations than in a client-server installations where the server handles much of the data processing.

    Considerations when choosing a network operating system:

    Regardless of HOW our organizations choose to network, it is clear that we SHOULD network. Only in this way will we be able to make the most of our equipment and communications infrastructure.

    2 - Wide Area Networks - WAN

    The various Wide Area Network Internet applications, such as Usenet and the World Wide Web, are discussed at length in our Cyberstrategy. While pre-Bronze Age machines are capable of accessing the more rudimentary elements of Internet, such as ftp and gopher, the more powerful implementations such as the World Wide Web are largely beyond their reach. And these machines have only limited potential for even Local Area Network integration at the enterprise level.

    Currently, the basic device for Wide Area Network access is the modem [MOdulator/DEModulator]. A modem transforms digital signals from a computer into analog signals for transmission over telephone lines, and converts incoming analog signals back into their digital equivalents. Initially, most modems operated at speeds of 1200 or 2400 bps [bits-per-second], though by the late 1980s units operating at 9600 bps were common. In the early 1990s modems operating at 14.4 kbps [thousands of bits per second] were introduced, and today's high-speed modems operate at 28.8 kilo-bits per second.

    Although contemporary 28.8 kbps modems are more than 20 times faster than modems of a decade ago, further advances in speed will require using a different technology. Modem speed is limited by the requirement to convert between the digital format of the computer and the analog format of standard telephone lines.

    Integrated Services Digital Network (ISDN) lines carry three digital channels: two "B" channels operating at 64,000 bps, and a "D" channel that carries control signals or serves as a third data channel, at 16,000 bps. With appropriate data compression schemes, ISDN can transfer data at hundreds of kilobits per second. ISDN has been under development for many years, but until recently it was joked that ISDN either stood from "I Still Don't Need" or "It Still Does Nothing" -- now this is changing.

    Even higher speed connectivity is available, and it may be anticipated that with time even ISDN will be superseded by faster leased-line connections. T-1, which carries data at 1,544,000 bits-per-second, could transfer a megabyte in less than 10 seconds. Although T-1 is the fastest speed typically used to connect LANs to the Internet, it is not fast enough for full-screen, full-motion video, which requires at least 10,000,000 bits-per-second. A T-3 leased-line connection carries data at 45,000,000 bits-per-second, which is more than enough for full-screen, full-motion video.

    ISDN, along with even higher speed WAN technologies [T-1, T-3 and SONET], will have a significant impact on LAN architectures as well. Currently, dial-up modems are connected to individual computer work-stations, bypassing an organization's local area network [if any]. The cost, complexity, and capabilities of ISDN and other high-speed WAN connectivity mandates routing through a client-server LAN -- individual client stations are connected to the LAN server, which is in turn connected to the ISDN gateway to the Internet. While peer-to-peer LANs were sufficient for small organizations in modem era, client-server LANs will be mandatory as we move to higher-speed implementations such as ISDN.

    In the past, our community's LAN implementations were largely driven by the distinctive organizational requirements. Some organizations with highly integrated staff functions moved quickly to implement LANs, while others with distributed office locations or decentralized operating styles found less utility in the LAN environment. Increasingly, however, the opportunities provided by high-speed wide area network Internet connectivity will move our entire community toward client-server LAN implementations.

    Robust Local Area Networks are increasing essential to the utilization of peripheral resources such as printers and scanners, and the effective coordination of staff work. Wide Area Networks, particularly the Internet, are increasingly essential to effective collaboration among organizations, between organizations and their memberships, and among the public at large. Emerging software implementations, such as the World Wide Web and groupware, will increase the central role of such networking over the next several years.







    Stations

    Tier Year Platform RAM Hard-Drive applications Speed Performance
    PC & Mac PC / Mac kb MB bits Mflops SPECint92
    Stone 1977 Selectric
    Apple II
    4 - 0
    8
    ? ? ? ?
    Zinc 1979 TRS-80
    Apple II+
    48 floppy 8 ? ?
    Iron 1981 8086/4-PC
    Apple II+
    64 floppy 8 0.0069 0.001
    Lead 1983 286/4-XT
    Apple IIe
    128 10 8 0.012 0.05
    Tin 1985 286/6-AT
    Macintosh
    256 20 16 0.033 0.1
    Brass 1987 386/16
    Mac II
    640 40 16 0.16 2
    Copper 1989 386/33
    Mac IIcx
    2,000 80 16 .27 9.4
    Bronze 1991 486/33
    Quadra 900
    4,000 120 16 .94 22.4
    Silver 1993 486/66
    Quadra 950
    8,000 300 16 1.2 39.6
    Electrum 1995 P5/100
    PowerMac
    16,000 1,000 16 & 32 12 120
    Gold 1997 P6/200
    PPC
    32,000 2,000 32 35 300
    Platinum 1999 P7/300
    PPC
    64,000 4,000 64 115 1,000
    Palladium 2001 P8/400
    PPC
    128,000 8,000 64 300 ? 2,500 ?
    Iridium 2003 P9/500
    PPC
    256,000 16,000 128 ? 700 ?? 6,250 ??

    Stations

    Tier Year Platform RAM Hard-Drive Op System Original Current
    PC & Mac PC / Mac kb MB price price
    Stone 1977 Selectric
    Apple II
    4
    System 2

    $3,100
    Zinc 1979 TRS-80
    Apple II+
    48 floppy CP/M
    System 3

    $2,595
    Iron 1981 8086/4-PC
    Apple II+
    64 floppy DOS 1.0
    System 3
    $5,200
    $2,000
    $29
    ? ?
    Lead 1983 286/4-XT
    Apple IIe
    128 10 DOS 2.0
    System 3
    $6,550
    $2,175
    $95
    $148
    Tin 1985 286/6-AT
    Macintosh
    256 20 DOS 3.0
    System 4
    $6,500
    $2,825
    $125
    $175
    Brass 1987 386/16
    Mac II
    640 40 DOS 3.3
    System 5
    $8,975
    $8,550
    $350
    $300
    Copper 1989 386/33
    Mac IIcx
    2,000 80 DOS 4.0
    System 6
    $5,100
    $5,050
    $350
    $400
    Bronze 1991 486/33
    Quadra 900
    4,000 120 Windows 3.0
    System 7
    $3,025
    $5,090
    $700
    $800
    Silver 1993 486/66
    Quadra 950
    8,000 300 Windows 3.1
    System 7 Pro
    $3,300
    $3,550
    $1,500
    $2,100
    Electrum 1995 P5/100
    PowerMac
    16,000 1,000 Windows 3.11
    System 7.5
    $3,695
    $5,309
    $3,695
    $5,309
    Gold 1997 P6/200
    PPC
    32,000 2,000 Nashville
    Copland
    $3,000
    $3,000
    $7,500
    Platinum 1999 P7/300
    PPC
    64,000 4,000 Memphis
    Gershwin
    $3,000
    $3,000
    $15,000
    Palladium 2001 P8/400
    PPC
    128,000 8,000 ? ?
    ? ?
    $3,000
    $3,000
    $50,000 ?
    Iridium 2003 P9/500
    PPC
    256,000 16,000 ? ? ?
    ? ? ?
    $3,000
    $3,000
    $250,000 ?

    Nets

    Tier Year Platform Internal External "push" "push" "interactive" "pull"
    LAN kbits/sec FAX E-mail Internet Website
    Stone 1977 Selectric
    Apple II
    Zinc 1979 TRS-80
    Apple II+
    Iron 1981 8086/4-PC
    Apple II+
    modem
    .3
    Lead 1983 286/-4XT
    Apple IIe
    .3
    Tin 1985 286/6-AT
    Macintosh

    AppleTalk
    modem
    1.2
    manual
    Brass 1987 386/16
    Mac II
    Novell 2.15
    AppleTalk
    modem
    2.4
    manual text
    elm
    Copper 1989 386/33
    Mac IIcx
    Novell 3.0
    AppleTalk 2
    modem
    9.6
    manual Eudora 1.3 ftp
    telnet
    Bronze 1991 486/33
    Quadra 900
    Novell 3.1
    AppleTalk 2
    modem
    14.4
    manual Eudora 1.4 ftp
    archie
    Silver 1993 486/66
    Quadra 950
    Novell 3.12
    AppleTalk 2
    modem
    14.4
    WinFax Eudora 1.4.2 listserv gopher
    veronica
    Electrum 1995 P5/100
    PowerMac
    Novell 4.1
    AppleTalk 2
    modem
    28.8
    WinFax Eudora 1.4.4 usenet
    Netscape
    WWW
    Netscape
    Gold 1997 P6/200
    PPC
    Novell 5.x
    AppleTalk x?
    ISDN
    128
    groupware groupware groupware
    Platinum 1999 P7/300
    PPC
    ? ?
    ? ?
    T-1
    1,544
    ? ? ? ? ? ?
    Palladium 2001 P8/600
    PPC
    ? ? ?
    ? ? ?
    T-3
    44,736
    ? ? ? ? ? ? ? ? ?
    Iridium 2003 P9
    PPC
    ? ? ? ?
    ? ? ? ?
    SONET
    155,530
    ? ? ? ? ? ? ? ? ? ? ? ?


    FAS Homepage | CyberStrategy Project | Search | Join FAS