NETWORKING AND DATA
COMMUNICATIONS BASICS

BY PAUL HOOPER AND JOHN PAGE

Networks are becoming very large. There are private networks with over 60,000 separate users. The City of New York leases 6,000 telephone lines for computer communications supporting services such as police, fire, and ambulance dispatch. Merrill Lynch & Co. links over 600 brokerage locations in the U.S., Canada, and overseas with a network that operates 24 hours a day, 7 days a week. DuPont has a network in the U.S. connecting 108,000 phones and computer terminals. The Health and Welfare Agency Data Center of California, which requires that 90% of online transactions be completed within 4 seconds, processes 6 million transactions per week with over 350 simultaneous users. Exhibit 1 provides more industry examples of networking.

Obstacles to Understanding Data Communications

The business user of data communications must have knowledge of the technical aspects of data communication in order to make the most of this tool. The computer industry is a long way from the type of standardization that is found in the consumer audio industry, where equipment such as tape and CD players, amplifiers, and speakers from different manufacturers and various countries are compatible. Standardized interconnection of communications hardware, software, and various computers is still a distant goal for a number of reasons:

Different Terminology. Various names exist for the same item. IBM, for example, often uses terms unlike those used elsewhere in the industry.

Different Approaches. While alternatives often exist to achieve an end result, none of them necessarily have any advantages over the others, and they all seem to remain in use. In the automotive industry, while Ford put the steering wheel on the left side of the car, Chevrolet conceivably could have put it on the right. That industry, however, recognized the value of standardization, which has become impossible in the computer industry, where there are many types of existing equipment, each using different approaches. For example, the connection to IBM's midrange computers has wiring, keyboard, screen, and control software that is different than the IBM mainframe connection.

Obsolete Technology. Even though no longer manufactured or sold, outdated technology is still being used and must still be understood. For example, while many manufacturers have replaced their former data transmission methods with superior ones, the former methods, now obsolete, remain in widespread use.

Mainframe Approach

Before the microcomputer emerged, the first mainframe-oriented approach to data communications and networks was developed by IBM for online transaction processing (OLTP). This approach dominates the current market, accounting for over 80% of high-end systems. There are more than 4,000,000 terminals connected to over 10,000 mainframes using this approach, which has two primary goals:

1. Having the mainframe do all processing of data, including retrieval of data from databases and computations.

2. Moving everything else off the mainframe. In practice this means that all communications-related tasks are delegated to other devices.

The most complete large-scale network design for mainframes is IBM's System Network Architecture (SNA). SNA is very flexible and it works. It is also comforting to know that the vendor is likely to be around in the future. The disadvantages are that it is extremely complex and enormously expensive, which makes SNA too difficult and costly for routine large-scale networks. Even though IBM introduced SNA in 1974, some important areas of the design are still in the planning stages. Its slow development underscores the complexity of data communications and networking in this environment.

As a result of SNA, IBM mainframe online processing is very efficient for transaction processing. As noted in the second primary goal above, other functions that could slow down the processor are moved to other devices. By providing centralized control over data, this helps to ensure security, accuracy, and privacy. As changes occur, however, it is difficult to modify the network with new devices and the movement of devices. Also, the mainframe approach, developed before the introduction of the microcomputer, does not take advantage of the capabilities of the microcomputer.

Minicomputer Approach

IBM mainframes were designed for transaction processing. Online nontransaction-processing applications, such as electronic mail, engineering computations, text processing, and system development are extremely inefficient and expensive on these systems. Digital Equipment Corporation (DEC) designed, and has been very successful with, their minicomputer, especially for these other nontransaction online applications. Because of its online orientation and low cost, the minicomputer can also do transaction processing, but on a smaller scale than the IBM mainframes.

A minicomputer can support a number of inexpensive terminals. Each minicomputer user has a terminal with a share of the central processor and its resources, such as tapes, disks, and printers. Minicomputer users can communicate easily; they are all connected to the same machine. They can share data and programs, such as inventory, programming, or engineering applications. Multiple copies of large programs and data files are unnecessary for multiple users.

Other firms followed DEC into the minicomputer market, producing proprietary operating systems for their computers. If a user wanted to change hardware from one company to another, it was also necessary to change software because the operating system changed. With the rapid changes in hardware, this situation eventually became intolerable to minicomputer users.

Ultimately, UNIX was created. It is the only alternative operating system to the proprietary operating systems of computer vendors. UNIX was designed at Bell Laboratories as a portable, efficient operating system for minicomputers, especially suited for
online, nontransaction processing applications. UNIX is now the dominant operating system for minicomputers. Only the DEC proprietary operating system is still being upgraded and sold. All other minicomputer vendors have committed to UNIX.

There is no established network standard for minicomputers comparable to SNA for mainframes. There are, instead, different methods of communication that provide needed services in different areas.

Transmission Control Protocol/Internet Protocol (TCP/IP). TCP/IP was developed for the Department of Defense (DoD) to create a network of major institutions that do DoD contract research. TCP/IP has been widely used in commercial applications because it is the only practical way for computers produced by different manufacturers to communicate.

TCP/IP allows people to use a remote computer as if it were local. This is more sophisticated than just sending messages between machines. TCP/IP is based upon packet switching to efficiently control an interactive session with a remote computer. In packet switching, the message is grouped into packets of approximately 2,000 characters which are then transmitted. If an error occurs, only the packet which was erroneously transmitted needs to be retransmitted. For example, you might think of a single line in a letter as a packet of information; the line makes little sense by itself, but makes complete sense when placed with all the other packets (lines) in the letter.

X.25. X.25 is an international standard that allows the widespread use of value-added networks (VANs), such as Telenet. A VAN is a publicly-available, packet-switched network that extends nationally and internationally to allow long distance data communication through the network rather than through phone lines. For example, if a computer user in one city needs to access a computer in another city, a local X.25 connection over phone lines makes it possible to connect the two computers. The alternatives would be to use the dial-up lines for a long distance call or to lease a line between the two cities. X.25 provides a more reliable connection than a long distance call and is less expensive than a leased line in all but the most high volume applications.

Exhibit 2 compares and contrasts the mainframe and minicomputer approaches to networking and data communications.

Microcomputer Approach--
Local Area Networks (LANs)

The microcomputer was designed initially as a personal computer. It is particularly good for personal, interactive applications (such as spreadsheets and budgeting), for personal data files such as mailing lists and employee evaluation information, and for word processing. Most microcomputers in business have been justified initially as personal devices for stand-alone applications. This is changing.

A LAN is a direct, high-speed connection of microcomputers that allows all the communication activities previously discussed. LANs are typically formed from a group of stand-alone microcomputers whose users need to communicate to efficiently carry out a task. In addition to microcomputers, a LAN typically requires the components described below.

Network Interface Cards (NICs). A network interface card plugs into an expansion slot on the microcomputer and gives the microcomputer a window to the network. The major types of network interface cards are token ring and Ethernet.

Token ring is a ring configuration. A token or message stating that the channel is clear is passed between microcomputers in sequence until someone wants to use the network. The microcomputer then takes the token and changes the message to busy and the busy message is transmitted. When the transmission is completed, the token message is changed back to clear and passed along. This token ring technique was derived from the SNA approach of mainframe computers.

Ethernet is a configuration which allows each microcomputer on the network to look at the line and see if it is busy. If it is not busy, then the computer is free to transmit its message. Two computers could try to transmit at the same time, creating a collision problem. The network design solves this problem by using a collision detection device. If the device detects another message on the line, it will retransmit the message after an interval of time. Ethernet is the primary non-IBM approach.

Connection Between NICs. It is necessary to have a connection between the network interface cards of individual microcomputers, tying them together. This connection is usually twisted pair or coaxial cable.

Network-Aware Software. To access the LAN, each microcomputer must run network-aware software that enables the microcomputer to access files or printer services available on the network.

File Server. The file server is a computer that contains the programs and data files that all microcomputers on the network share. This is typically the largest microcomputer on the network, or it could be a minicomputer or a mainframe.

Network Operating System. The network operating system allows the file server to be utilized by many different microcomputers simultaneously. A network operating system runs only on the network's file server. Most often, the network operating system is Netware, from Novell.

Generally, these different components can be changed without changing the other components. For example, both token ring and Ethernet can work with Netware because the choice of network interface card is separate from the choice of network operating system.

Advantages and Disadvantages of LANs. The microcomputer is growing in capability. Microcomputers operating on a LAN often replace minicomputers. With a large file server, perhaps even a minicomputer, the LAN is beginning to provide very good small-scale transaction processing
capabilities.

Over the years, traditional applications for minicomputers and LANs have developed. The decision of which to use depends primarily on how much of the data and information on the system are shared. The more information shared, the more attractive a minicomputer; while the more data that is distributed with each user having his or her own data, the more attractive a LAN. For example, an inventory system is best done on a minicomputer because every user will share the same information on parts and their availability. At the other extreme, managers who each have information on their particular employees' efficiency are best served by a LAN.

Microcomputers and SNA. There is no need for a user, even in a mainframe environment, to have both a microcomputer and a terminal. The microcomputer can do stand-alone processing, such as budgeting, word processing, and personal data bases, and can also emulate a terminal. An emulation board converts between the Extended Binary Coded Decimal Interchange Code (EBCDIC) of an IBM mainframe and the American Standard Code for Information Interchange (ASCII) of the microcomputer; so typically the microcomputer displays the same screen as the mainframe terminal.

Exhibit 3 summarizes the important characteristics of the mainframe, minicomputer, and microcomputer approaches to networks and data communication.

Electronic Data Interchange (EDI)

Many businesses have recognized that there is wasted effort when both their company and its suppliers or customers have computer systems without having data communications among the systems. For example, a company may use a computer to prepare a purchase order. The supplier, upon receiving the order, has to enter the information into its computer system. Reentry of data into the computer system also will be done when the supplier sends an invoice to the company, when the company receives a purchase order from its customer, and when the company sends an invoice to its customer. A strong incentive exists to send the data to and from the supplier, the company, and the customer electronically without printout, reentry, mail delays, and mailing expense. The term used for this process is Electronic Data Interchange (EDI).

Initial EDI systems developed by K-Mart and American Hospital Supply Corporation (AHSC) were very successful and widely publicized. Many firms, such as Haggar, followed their lead. Retailers of Haggar clothes now read bar code information and transmit sales data to the Haggar mainframe. Haggar is also linked to its fabric suppliers in a similar way. Many firms, such as Sears, now require that their suppliers use EDI.

Because EDI systems were developed independently, the result was a wide variety of formats which did not communicate with each other. This would be similar to having AT&T telephone customers unable to call MCI or Sprint customers. Eventually, a standard was developed by the American National Standards Institute (ANSI). Firms such as Procter & Gamble led the move toward standards because of their need to constantly deal with so many other companies.

X.400. None of the EDI standards were complete solutions to data communication problems because special EDI connections required between companies were difficult to arrange. Business needed a standard way to use public or private networks for EDI and other forms of communication. A response to this need is the X.400 standard. Any network that corresponds to the X.400 standard can communicate with other X.400 systems and transmit telex, electronic mail, formatted documents, electronic data interchange, videotex, and voice as well as other types of communications. Vendors currently using the X.400 approach include General Electric Information Services, IBM, and Control Data.

Electronic Mail. The movement to X.400 enjoyed a major push from the users of electronic mail. The major electronic mail service providers, such as MCI mail and Western Union's EasyLink do not have compatible systems. A user signed on to one provider could not send an electronic mail message to the subscriber of another system. Gradually, LANs, corporate networks, and electronic mail service providers are adding X.400 capabilities.

Just in Time Inventory (JIT). One goal of many companies is to have a JIT flow of inventory. Long-term arrangements can be made with one or more suppliers to deliver items only when necessary for production. Eventually, this can reduce inventory quantities from several months or weeks on hand to as little as a few hours depending on the distance of a supplier.

Electronic Funds Transfer (EFT). Funds of all sorts can also be transferred electronically. The Federal Reserve System operates a wire transfer capability for banks. A similar process is now being used by all types of organizations. For example, employees' pay can be deposited directly into their bank accounts, and funds to pay suppliers can be transferred electronically to suppliers' accounts. This approach eliminates some internal control problems, as well as printing and mailing checks, but it adds the internal control problem of limiting access to the EFT capability.

Systems Application Architecture (SAA)

The future direction of the IBM approach to data communication is away from the centralized control of the mainframe and toward a network of mainframes and microcomputers, as peers. This peer-to-peer networking is growing, but it requires a more powerful microcomputer than typically available to microcomputer users. Part of the IBM movement to peer-to-peer networking is SAA. The purpose of SAA is to make certain characteristics consistent among the different IBM computers. SAA has defined standard ways for all systems to communicate with users, software developers, and other systems. For example, each program will use F1 as the Help Key and each menu bar will begin with File as the first choice. SAA began life as a mainframe-dominated specification because the focus was on program portability, so programs could be written on one type of computer and moved to another. However, future versions of SAA will rely more on microcomputer networks to better distribute the needed business processing.

Open Systems Interconnect (OSI)

When a network spreads over a large geographic area, it is often called a wide area network (WAN). Many WANs consist of different types of computers connected worldwide. There are three available network standards that can provide the backbone to sustain such a diverse network. Each of these different standards define how data is formatted, transferred, routed, and retransmitted if errors occur. These definitions, in turn, are implemented by connectivity software products incorporated into operating systems and built into communications hardware.

Two of these standards, SNA and TCP/IP, were, in part, discussed earlier. SNA has become a de facto industry standard by virtue of IBM's control of 80% of the mainframe market. TCP/IP is a series of specifications originally developed to ensure interoperability among U.S. Defense Department networks. It has since found widespread commercial applications. Unfortunately, TCP/IP is incompatible with SNA, and is only a partial network solution.

The third standard, OSI, which is of emerging importance, is a model developed by the International Standards Organization (ISO) for use in designing multivendor networks. Though the OSI model has not yet been completely implemented, parts of it are operating successfully. As OSI becomes more complete, TCP/IP will become less significant. However, because TCP/IP is
currently working and inexpensive, in the near-term it is indispensable to most
businesses.

The network model most suitable for the communication and integration of systems from different computer vendors is OSI. The OSI model divides communications into seven layers and has specifications for each. These specifications are functional only and the methods to be used are left to the designer. Although OSI products have been very slow in coming, those products that have appeared are open--anyone can use them and connect to all other users of OSI products. A significant drawback of OSI is that most currently installed networks were developed before the OSI model and are inconsistent with it. However, if data communications and networks are to achieve their potential, products incorporating all levels of the OSI model will probably have to be a part of the approach.

Exhibit 4 lays out the important basic choices a company faces in creating and implementing a networked, data communications environment. These choices must be resolved regardless of the approach--mainframe, minicomputer, microcomputer.

Paul Hooper, PhD, is a professor of accounting at the University of Delaware. John Page, PhD, CPA, is an associate professor of accounting at Tulane University.

A WIDE RANGE OF CHOICES--MORE STANDARDIZATION IS NEEDED.IN BRIEF

Three Ways to Connect

Networks reduce costs, provide better customer service, and improve managerial decision making. The authors present an overview of the basics of what a business or accounting manager should know in order to understand and use networks and data communications effectively. The prime focus of the article is directed at three basic approaches to business networks: mainframe, minicomputer, and microcomputer. A manager could encounter all three in various levels of activity. They address the pluses and minuses of the various systems and offer up exhibits, three of which demonstrate items such as options available, comparisons, and advantages and disadvantages. They also discuss the obstacles to understanding these tools, created in a large part by the computer industry itself.