Business-networking objectives

INTRODUCTION A major corporation headquartered in Houston conducted a one-day meeting. The meeting included all of the key managers and executives in the company who had responsibility for operating or using the corporation's computer networks. About two dozen executives and high-level managers in the room, with widely diverse backgrounds and perspectives, attended. The meeting agenda had only one item on it: Business Network Objectives. The goal of the meeting was to agree upon what the objectives of the corporation's network management team should be. By the end of the day, the following list had emerged:

Satisfy current needs for services in a cost-effective, systematic manner.

Position the organization to meet future network-service requirements without costly redesign or unnecessary user disruption.

Achieve a balance between shorter-- term and longer-term objectives to help control business risk resulting from network change.

Establish user profiles for computer applications that help to identify strategic and tactical requirements for the network - that is, those requirements that go beyond routine operational needs.

Construct service-level agreements between network management and users that include performance measures that are meaningful in terms of the users' work environments and are based upon an understanding of costs for different service levels.

Maintain transparency of the network from the users' perspectives as modifications are applied over time.

Establish techniques for anticipating changes in network requirements and for predicting the effects of network change.

Standardize network specifications and configurations to the extent possible within limits of user requirements, cost, and management strategy.

Establish a network design that facilitates support and service of networking equipment with an architecture that provides for centralized network management and control whenever feasible.

Establish policies to control and minimize the complexity of the network and to instill ease of maintenance as a selection criterion for choosing equipment.

Achieve interconnection flexibility for network components and subnetworks that may be specialized to particular application areas.

Develop and implement cost versus benefit analyses to support acquisition of networking equipment.

Establish a standardized primary network backbone to interconnect with subnetworks as required.

Integrate network-usage files with other files for computer-use reporting, capacity analysis, and performance analysis into a single consistent database architecture.

Achieve a flexible networking environment that can be easily upgraded when needed and that provides users with consistent, uncomplicated interfaces to applications systems as needed.

This list hits the "nail right on the head." It is a good generic list of network-management objectives. But how does one put such a list into effect, ensuring that these objectives are met with any reasonable level of consistency and regularity? This goal is not easy. Today, even large corporations tend to get bogged down with outdated network-management tools and techniques, or have well-meaning, but often inadequately trained network analysts who rely mostly on intuition, guesswork, and trial and error. Achieving the goals outlined above means moving network management away from heuristics and toward a more systematic approach, away from a reactive reliance on educated guesswork and toward active reliance on sound principles of business management.

TRYING HARD TO BE LUCKY

Ever hear the phrase, "Sometimes it is better to be lucky than to be good"? Managers of information technology are well aware of this old adage, especially when it comes to computer networks. Most of the established corporate computer networks began their existence in the 1960s, evolved in the 1970s, expanded rapidly in the 1980s, exploded during the 1990s, and there is no end in sight. Not only is the demand for networking frequently outstripping corporate ability to provide the services needed, but also the underlying network technology itself is in an ongoing turmoil. Just think about the string of buzzwords in this field, each of which signals a revolutionary change in the technological context of network architectures: Distributed Processing, Packet Switching, Digital Transmission, Local Area Networks, Internet, Intranet, Extranet, World Wide Web, Fiber Optics, Satellites, Peer-to-Peer, Broadband, Switches, Exchanges, Client/Server, Firewall, Thin Clients, Wireless, and on and on. With the technology increasing in complexity and constantly churning, little wonder that network-management teams put a premium on "just being lucky." It is part of the game.

NETWORKING CHAOS Entropy is the tendency for a system of any kind to slip into a state of chaos if left unattended, and networks are supremely entropic. Even a network with excellent performance, if not managed carefully, will quickly slide downhill into chaos and disarray. Network managers can never really relax and assume that a favorable status quo will be sustained for very long.

One major reason why vigilance is required is that user demand for available network services is not static. It shifts, for example, as business and economic situations change, or as users become better trained to use systems more fully, or as existing applications are gradually accepted and used by more departments in a company. This shift usually occurs day by day and happens at different rates for different parts of the network. Over a year, or even a few months in some cases, this gradual shift can have a profound influence on the performance of the network. It means that the network must be periodically redesigned even if the networking infrastructure and the basic services provided do not change significantly.

The other major reason why vigilance is needed is that new requirements for network services frequently surface, causing the existing network configuration to need modifications (for example, to add new application systems or to make the network faster or more secure). Network upgrades include configuration changes such as adding terminals, increasing the number or speed of the lines, upgrading hardware or software on processor nodes, restructuring the servers' capacities or functionality, and so on. The goal from a network-management perspective is to implement the new services successfully while minimizing any effects of the upgrades on the users of the previously available network services. Unfortunately, it is often difficult to anticipate how changing one part of the corporate network will affect the performance of services in other parts of the network. That is, a change in one part of the network can cause unanticipated bottlenecks in other parts. These kinds of situations can wreak havoc with network-management plans.

Together, these two factors (entropy and the pressure to constantly upgrade) over time render the best static network configurations ineffective. Imagine the impact of these factors upon less-well-designed or poorly controlled networks. It is chaotic. Network efficiency and effectiveness are moving targets that the network manager must continually try to hit. Managing network growth means identifying and dealing with both the shifts in user demand associated with an installed network-baseline architecture, and with the increases in demand represented by the addition of new services on the network. It is a complex proposition.

A DELIVERY-SYSTEM PERSPECTIVE

One of the key objectives of systems analysis and planning for information systems is to ensure that adequate processing capacity exists for the "delivery system" - that is, for the combined computing and networking portions of the information-- systems environment taken as a whole. An information-systems manager can add only so many new users or new services before degrading the overall quality of service provided across a network, because the system simply gets overloaded. From the users' point of view, whether the computer is "out of gas" or the network is hammered, "the whole damn thing is broken." If demand for either computing or networking services exceeds available capacity, then the overall quality of service for information-systems users declines rapidly. And response time becomes intolerably slow or the network stops working altogether.

Thus, computer- and network-- capacity planning are closely linked. Essentially, they create different views (using different measures) of the same workload on an information system. At a conceptual level, network analysis and planning is an adaptation of established computer-systems analysis procedures that deal with such issues as user-workload evaluation and computer-systems capacity management. Unfortunately, while the computer-- systems capacity management issues are fairly well understood in corporate circles, the equivalent problem in network management is much less understood, or even recognized.

This is a particularly thorny problem. When there are difficulties in the network, the situation often goes unrecognized until the user begins to experience degraded service. The user does not know whether it is a computer or a network problem, a temporary inconvenience or a symptom of a more permanent malady. Furthermore, by the time that the delivery-system service levels begin to decline, it is often too late to avoid increasingly serious degradation of service before the problem can be corrected. The reason for this situation is that network redesign and modifications to fix problems require too much lead time. Old equipment usually must be replaced with new equipment that must be ordered, installed, tested, and put into production status. This process may take weeks, or more; in the meantime, the user is struggling to operate in a degraded systems environment and usually is complaining a lot to turn up the heat on the network management team.

Managing network resources effectively in a typically pressure-- packed, totally reactive mode usually does not work. The problems are too complex, too interrelated with other services, and too visible to the user organizations to justify relying on a "seat-of-the-pants" management style. Therefore, the network management process should include procedures for anticipating the changes that will be needed in the future before they become problems. The basic tool in this process is systematically forecasting underlying changes in workload characteristics, evaluating the effects of these changes on the network-configuration needs, and then upgrading the network before problems arise. This is easily said, but it is not easy to do accurately and consistently.

NETWORK MANAGEMENT

The overriding goal in network management is to design and maintain an information-systems networking infrastructure that is both flexible and cost-effective. Given the way network requirements change over time, flexibility is crucial in network design. In most cases, this objective translates into a requirement to design networks with services that are technically compatible with one another and, to the extent needed and plausible, that are effectively integrated to facilitate sharing of resources. It also means understanding specific user requirements and being prepared to adapt as users' needs shift to reflect changing business priorities and economic situations.

These factors imply requirements to continuously monitor network utilization, to forecast future network activity, to anticipate shifts in network usage before they happen, and to reconfigure the network before changes in usage patterns cause the performance of the network to degrade seriously. The process of increasing or decreasing the capacity of appropriate network components to reflect changing levels of demand for network services should be undertaken before the growth in demand actually overtakes the capacity of networks or subnetworks in the system. Otherwise, network management cannot avoid service degradation unless it is very lucky.

The need to anticipate changing network demand means that various trends in network usage must be evaluated to forecast shifting workloads. In a complex networking environment, this forecasting is most reliably accomplished by the use of statistical models. Forecasting models employ historical usage statistics for each line in the network and can forecast usage trends to help identify bottlenecks in the configuration. The resulting forecasts are then used to help design and implement successive generations of network configurations as new components are added to or old components are removed from the network as services are upgraded.

CONCLUSION

There is little doubt that the modern corporation depends upon computer networks to a degree that has never before been seen. That dependency is ever-increasing, and it means that corporate management must first recognize and then address the special range of problems implied by trying to effectively manage complex, unstable network architectures. Realistic goals and strategies, like the framework of objectives of the Houston firm described above, and related mechanisms for monitoring and controlling network performance are critical to the successful management of the modern corporate networking infrastructure. These networks pose for businesses both a challenge and an opportunity. The challenge is to manage all this complexity effectively, and the opportunity is to reap the benefits of being well connected in a complex and increasingly wired world economy.

Charles K. Davis is a Professor in the Cameron School of Business at the University of St. Thomas in Houston, Texas, and is a past Chapter President of Phi Kappa Phi. He has taught previously at the University of Houston and held analyst and management positions with IBM, Chase Manhattan Bank, Occidental Petroleum, Pullman Incorporated, and Deloitte & Touche.

Copyright National Forum: Phi Kappa Phi Journal Summer 2001
Provided by ProQuest Information and Learning Company. All rights Reserved