Technical Advances in Data Communications Alter Large-Scale Messaging Environment.
Our purpose in this article is to discuss the reasons for the changes, to describe fouir models of message networks that are becoming commonplace with today's technology and to identify the key issues in designing a large-scale messaging network.
With the turmoil created by recent, rapid technical advances, a period of re-appraisal has clearly been in order. It catalyzed a search for solutions that is still going on. The problem is one of abundance. Even in small business organizations today, computing and communications power is virtually within arm's length, in the form of word processors, personal computers and various types of I/O terminals.
Only five years ago, data was data, word processing was word processing and communications, communications. Each was a separate function with unique equipment, specialized procedures and dedicated staff. Talk was of crossover software, multifunctionality of DP/WP terminals and the difficulty of achieving these objectives. Message communications was not considered "in the mix." Isolated from it all, it was a world all its own.
That is no longer true. Now, management recognizes that the same powerful forces that altered the nature of accounting, finance and general decision-making, can do the same for communications. It is now seen that corporate-messaging systems can be developed that are vastly more efficient and cost effective than the mail, couriers and the telephone.
While all of the problems have not yet been solved, there has now been a significant blending of communications and computer technologies. Corporate management information departments, for example, are well aware of the incursion of personal computers and are hard at work integrating them into the corporate EDP and communications networks.
The message communications function, too, has graduated, and is increasingly represented by sophisticated multifunction, multiport, microprocessor-based terminals that are, in effect, dedicated microcomputers.
Advances Can Cause Problems
Such technological advances have also generated problems for organizations with a diverse hardware and software base. Much of it is incompatible, and for these individual units to communicate, methods must be found to facilitate their access to each other and to equipment at other levels of the network structure. Software is usually the key to successfull network integration, but unfortunately, it lags far behind the currently available hardware.
The same situation exists when creating large-scale message networks that involves carriers that are not, of themselves, compatible. To be effective, the network must make them appear to be. This is a minor problem when dealing with the more widely used record carriers, but it gains dimensions as other possibilities are exploited, such as message traffic on data lines, specialized and dedicated private lines and so on.
A comparable situation could be found in a railroad system in which each competing carrier has built a line with track of a different gauge. Interline movement of goods is impossible until a means is found to transfer them easily from one carrier to another.
Exploiting Large-Scale Network
Nevertheless, the simple fact that potential for developing large-scale communications networks exists virtually mandates that it be exploited. And this is already happening in many organizations. The cumulative power of the PCs, message communications terminals, word processors and various dumb and smart DP terminals in today's offices is impressive--and it begs to be used.
The fact is that means to achieve large-scale, wide-area networking exists, it is often already installed and it is normally to the corporate advantage to facilitate its development.
The foregoing observations are based on work now being done and, in some cases, already completed, by Sidereal Corporation's customers who have successfully employed the company's message communications terminals and switches as fundamental components of the kinds of networks under discussion.
The rest of this article will focus on hypothetical message network models similar to systems now in use by Sidereal customers. We believe these types of network can be implemented successfully by many organizations within a reasonable cost range and with minimal modification.
The objective of a network, in most cases, has been to find solutions that grant maximum flexibility in internal communications--often a form of electronic mail--and maintain a full-scale, unimpeded capability for communicating over all the standard domestic and international message carriers as well as other private or specialized networks.
Interfacing Separate Networks.
The network models that follow include the public network, the dedicated message network, the data processing network and the combination LAN/WAN (wide-area network). These are not mutually exclusive and are often found coexisting in the same organization. In fact, a major issue facing network planners and managers is how to interface these heretofore separate networks.
Public network make use of existing services provided by "public" companies--that is, those providing services to all willing to pay. These service providers include the IRCs (RCA, MCI, FTCC, Western Union, TRT and ITT), AT&T and other DDD or DDS providers, and value-added networks such as Tymnet, Telenet, Geisco and CSC. This type of messaging network is common in smaller firms or where the messaging volume is too small to justify major investments in sophisticated, customized messaging systems. The investment is in the terminals or interface equipment needed to communicate with the public network.
Implementing a large-scale network using existing public networks involves creative management of message traffic to minimize line and tariff charges, and the use of equipment with power and features that complement the management effort.
As the saying goes, it does not take a sledge hammer to drive a tack. The power of each office's terminal or system should relate to its message requirements.
An Actual Situation
An actual situation finds a headquarters in New York, with major international offices in London, Singapore and Dubai. This company routes as much international traffic as possible through London, with switching performed at two levels: at the user's hub terminals and through the public network's switches.
Multiport Micronet 8 terminals are located in the home office and in the London office, connected by a dedicated line operating at 4800 baud to provide large-scale high-speed message transmission capability. New York automatically feeds out to domestic destinations, with London handling the various overseas post, using the basic telex services.
The respective branch sites are equipped with whatever level of terminal they require, ranging from one or two-line models up to fairly sophisticated desktop terminals such as the four-port Micronet 25.
This configuration represents a world-wide network that requires for its engineering only the intelligent selection and use of terminals. It provides excellent service to organizations whose activities involve far-flung but not necessarily complex-message communications.
A dedicated message network finds communications switches positioned at key offices throughout the organization's operating area, domestic and international. Size of the switches ranges from 24 to 500 or more ports, depending upon a switch's assignment within the hierarchy. This network of centralized and dispersed store-and-forward switches is able to accommodate a rather large selection of protocols, codes and transmission speeds, permitting great latitude in carriers with mainframes and minicomputers can be performed in batch models.
As a case in point, the Micronet Message Switch, with up to 24 ports, accomodates some 30 protocols including Telex I, Telex II and DDD 110/300/1200 asynchronous, HDLC; polled network protocols such as 8A1, 83B3, and 85A1; TTY, X-on and X-off; batch protocols such as IBM 3270 (batch) and i780/3780 bi-synchronous, and Autodin and Infocom ClassI, II and III, among others. Many other specialized protocols can be accommodated through software modifications. Necessary code, speed and protocol conversions are performed automatically.
As such, this type of switch is suitable for master switching in medium-volume wide-area dedicated message network is suitable where volume and diversity of traffic on the network can justify the required investment in hardware, software and manpower. Other reasons for a dedicated system include security, centralized control, flexibility and speed in reconfiguration.
While relative few organizations in the past were able to afford the minimum investment of about $500,000 to establish this type of system, the advent of micro processor-based systems has brought the cost down considerably. We expect that many smaller corporations will now be able to profit from this technology.
In this model, the corporate mainframe is the focal point and serves as the primary message switch and mail repository. The objective is to capitalize on an already existing capability in data communications and create a comprehensive electronic mail system by using the blend of PCs, data terminals and message terminals already in place within the organization.
There are several popular software packages available for mainframes that turn them into message network hubs. IBM's PROFS, DEC's Decmail and Wang's Mailway are three examples of such packages.
The store-and-forward message switch can serve a number of purposes in this configuration. It serves as the communications fron-end to the mainframe, batching and addressing messages prior to their entering the mainframe for distribution to mailboxes or remote terminals. Or, it can serve as the gateway for all outbound message traffic to domestic and international record carriers, with the mainframe automatically handing off such traffic to the appropriate switch.
Switches at dispersed locations serve as the electronic mailbox or message distribution coordinator for the local site--or for several related sites.
Switches Serve as "Glue Boxes?
Switches can also serve as "glue-boxes" that patch together the diverse mainframes and individual terminals or PCs through automatic code, speed and protocol conversion--in effect, making each look like the other to the message network.
An extremely comprehensive network for both traditional message communications and electronic mail can be built by combining the approaches of the wide-area message network with a local-area network for internal distribution, especially where PCs are becoming important to an organization.
For example, assume that a main office in Atlanta has 150 PCs that are linked by one or more LANs. Further assume offices in Chicago, Dallas and New York with 40, 55 and 80 PCs or similar devices that have been networked. The LANs themselves now can be networked, using an HDLC protocol with a message switch. The switch performs all of the store-and-forward functions, including serving as the electronic mailbox, with nothing required beyond dial-up and password entry.
Message generations is similarly uncomplex. Each application requires its own distribution and access system, to be sure, but once this is set, electronic mail is probably easier than traditional postal mail. For external message traffic--TWX, TLX, DDD--the text can be prepared on the PC, just as it if were a dedicated communication workstation, and transmitted to and through the message switch to the designated destination. An option is to have external traffic further dispatched by the switch to a central message communications center at each location.
More Than One Switch May be Desirable
Depending upon the size and geographic spread of the organization, more than one switch may be desirable, and positioning of the switches may relate to communications volume and line costs. The extreme flexibility of system configurability using powerful, low-cost, microprocessor-based message switches makes it possible to build highly functional networks at relatively low cost. A further advantage of having multiple-switch systems is that of redundancy, each has sufficient power that if one is down temporarily, another is fully able to take up the slack.
There is most definitely a message to be found in the various combinations of data/message processing technology. It is that a genuinely new area has evolved in management of such systems. There really is no legitimate separation of information functions anymore. Certainly, there are and will continue to be necessary disciplinary specializations, with lines of responsibility and authority clearly drawn.
But now, communications and data processing are so closely related that one seemingly cannong function at peak level without the other. The two management functions have much to give to each other, and between them, they will probably develop the high-performance engine needed to power today's information-driven organizations.
In implementing a large-scale network, the manager must take into account several key issues that will control the success or failure of the endeavor. These issues are:
* Preservation of existing investments in hardware, software and networks. In many cases, any large-scale network proposal will have to incorporate an organization's existing base of hardware and software and, in most cases, some kind of network. This may include a data processing environment of on-line, interactive sessions; a local-area network of word processing machines; and some kind of corporate data processing philosophy (distributed versus centralized control). In addition, an existing network architecture such as SNA, DECnet or Ethernet may have to be incorporated into any large-scale network proposal using customer software and hardware emulators.
* Flexibility to incorporate new technologies and carrier services. any large-scale network design will have to be flexible enough to incorporate future advances in technologies and carrier-provided services. As costs go down and enhancements are implemented, the large-scale network user will want to take advantage of them.
* Flexibility to incorporate new protocols and equipment. Chances are than a wide variety of equipment will have to be connected to this large-scale network, and once implemented, it must be flexible to accommodate new protocols, interfaces and equipment. A variety of protocols (async, bisync, SDLC, HDLC, 8A1, 83B3, X.25 ...), electrical interfaces (RS-232-C, RS-441, current loop...), and information types (data processing, facsimile, telex, word processing, OCR input, barcode processing...) will all have to be considered in any network design effort.
* Costs of hardware, software, communications facilities. Cost may be one of the most important underlying factors in any large-scale network design. Any hardware/software/communication facility solution may prove too costly to implement and support. In such cases, a less comprehensive and correspondingly less expensive solution that incorporates key portions of an all-encompassing design may prove acceptable.
Sensitivity to Audience
* Transparency. Any network design, whether large-scale or not, must be sensitive to the audience it supports. A network design that has been implemented to appear transparent to the end user will be more readily accepted than one that disrupts a person's day-to-day routine.
* Information flow within the network. In designing a network, an engineer will analyze various types of information with regard to the originator and final destination, the urgency of delivery (immediate, normal, deferred), the integrity of the information (special control characters in word processing documents, transparent data to a data processing host CPU, packed or binary information), and any security associated with this type of information (user password, encryption).
* Expandability. Especially with proliferation of PCs in the contemporary organization, network planners need to consider growing demand to bring new, unsophisticated users into the networks. Further, the current variety and lack of sophisticated communications capabilities on PCs dictates flexibility in accommodating such new users.
Traffic Analysis within a Network
* Traffic analysis within a network design. The engineer must consider peak information flow during any given hour, maximum information within a 24-hour period (day), and average size of each information type. These inputs are used to analyze any kind of queuing the network might experience under a fully loaded situation (peak hour) and under normal conditions (24-hour time period).
The activities of companies and institutions are becoming increasingly communications-intensive, hand-in-glove with rapid growth, diversification and geographic dispersion. Comprehensive, affordable networks are one of the vehicles that not only sustain these organizations, but also make feasible the ambitions and aspirations for the years ahead. It is not too much to say that message communications are the cement that bonds the pieces of today's enterprise. This being so, it follows that for engineering and outfitting a network, too little is potentially far more costly than too much.
|Printer friendly Cite/link Email Feedback|
|Date:||May 1, 1985|
|Previous Article:||Contention Control Critical in Mobile Radio Datacomm System.|
|Next Article:||Demand for Information Spurs Growth of Data Communications.|