Printer Friendly

Use modern approach to network integration.


The needs of a modern enterprise are quite sophisticated. Requirements for traditional voice communications were met by a corporate voice network based on a distributed architecture. (See Figure 1.)

The needs for traditional data communications and information processing were met by a centralized set of mainframes and by a large number of distributed dumb terminals. A low-speed hierarchical data network based on interconnected multidrop lines provided the communication between the dumb terminals and the hosts. (See Figure 2.)

During the last decade, the emergence of digital switches, digital T1 transmission lines, and T1 multiplexers brought about some integration at the backbone network level. (See Figure 3.)

But the recent rapid proliferation of PCs, intelligent workstations, and departmental LANs is challenging telecommunications managers and information officers to model, analyze, and design integrated multivendor enterprise networks, each capable of handling a myriad multimedia applications in a cost-effective manner.

Recent progress in technology and standards development is providing a large number of building blocks such as bridges, routers, gateways, privately owned digital microwave and fiber facilities, intelligent T1/T3 multiplexers, common-carrier-provided interfaces for X.25, and frame-relay for synthesizing national and global enterprise networks.

Dizzying Possibilities

Consequently, the number of possible solutions for enterprise networks is increasing exponentially while existing design tools become inadequate. There is a definite need for a new design methodology that will allow the corporations to model, analyze, design, and plan cost-effective integrated multivendor networks interactively and iteratively.

The traditional design methodology was heavily dependent upon a mainframe. Each network was modeled, analyzed, and designed with one specific application (e.g. voice, data, video, etc.) in mind. Each such network required a large database of traffic tapes and FCC-regulated tariffs.

Each set of design parameters was sent to the mainframe operating in the batch-processing mode.

The traffic tapes were analyzed for hourly traffic profiles and traffic flows between all locations, telephone numbers were repeatedly translated into V&H coordinates, and a network was synthesized for a specified set of switching/multiplexer nodes during computation of monthly costs for all tariffs.

This effort generally resulted in a large output transmitted overnight to avoid congestion of leased lines and/or reduce transmission costs.

Since no one has ever seen such a thing as an ultimate software package that provides an optimum network at the first try, the designer was forced to attempt only a single "what if" solution each day.

Designer productivity was extremely low.

Since the task of modeling and designing integrated enterprise networks will require an extremely large number of "what if" solutions, the traditional approach becomes almost useless.

The need for a user-friendly interactive approach is critical.

The Solution

The new methodology must employ a modern desktop workstation and a very user-friendly interface to achieve the interactive process.

This goal can only be achieved if one harnesses the following rather obvious but not very well known truths:

* No new topologies. LANs and MANs can be modeled using star, bus, or loop topologies. WANs can be represented by interconnected star, multidrop, and mesh topologies. This set of useful topologies has remained constant despite the great strides that have taken places in the fields of data processing and telecommunications during the last two decades.

There is, however, one major difference. The bandwidth on major routes can now be shared by most applications through such technologies as T1/T3 multiplexers, fast packet switching, voice packetization, and frame/cell relay.

In the resulting WAN topologies, voice LANs (e.g. PABXs), data LANs, and lower-level tandem switches can now all be treated as concentrators of bandwidth demands.

Another major change has to do with increased complexity in computing end-to-end delays and LAN/WAN throughputs within the umbrella of peer-to-peer communication. But this represents only an increase in analytical complexity and has nothing to do with network topologies.

By focusing on network topologies, one can really accelerate the network design process.

* Network architectures and design parameters--not tariffs--are what influence topology. Astute network designers in the past always observed that as long as the cost of any given link type increases with distance, network topology remains unchanged, even when tariffs are changed (e.g. when different common/specialized carriers are selected).

consequently, while accelerating the design process, one needs only a small set of simple tariffs for the synthesis of even a large network.

Conversely, one does not need to maintain a very large database of all tariffs from all the common or specialized carriers during the process of topology optimization.

* The designer does not need a truckload of traffic tapes. The author never forgot the day when he received a truckload of traffic tapes from a client. The tapes represented a month of call logs from each of the 300 locations of the corporation. The entire data that were ultimately reduced into busy-hour statistics covered only a few pages.

The transmission facilities are always sized according to the BHR traffic intensities (in units of Erlangs or BPS), and the switching nodes are sized to handle about twice the throughputs required during the busy hours.

The design process for an optimum network topology can be really accelerated if the BHR statistics (in the form of aggregate nodal traffic intensities or from-to traffic flows) are employed. If other resources (such as number of agents for an ACD) must be computed for all hours or days, one can employ a daily/weekly traffic profile. The BHR statistics and daily/weekly traffic profiles can be derived from the traffic tapes or summary traffic reports collected at each node on an off-line basis.

* Integer arithmetic accelerates the design process. By defining most of the often-used quantities--such as traffic loads as long-integer (or 32-bit equivalent) numbers and avoiding the use of real numbers as much as possible--one can accelerate the network topology optimization process significantly.

Furthermore, by defining the V&H coordinates for each node at the outset, one can avoid repeated translation of telephone numbers into equivalent V&H coordinates during the design process.

Several software packages harnessing one or more of the above truths have started to appear in the marketplace. Expect additional progress in the very near future.
COPYRIGHT 1991 Nelson Publishing
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Sharma, Roshan L.
Publication:Communications News
Article Type:tutorial
Date:Jan 1, 1991
Previous Article:Voice systems do creative work for users.
Next Article:Lawyers learn how to upgrade at 'Network U.'

Related Articles
The use of CAI for distance teaching in the formulation of search strategies *.
SMT/HYBRID/PACKAGING 2002 to arrive in Germany.
Teamwork needed for Net Centric Ops, AFEI head says.
An integrated framework used to increase preservice teacher NETS-T ability.
O'Reilly & Associates.
O'Reilly & Associates.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters