Previous Table of Contents Next


NETWORK SYSTEM PERFORMANCE

Many network systems are implemented without any regard to performance. As long as they satisfy the basic needs for communications, everyone is happy. Often, no effort is expended in:

  Predicting or measuring the actual performance of the system
  Making any measured systemwide improvements after the system is operational

The lack of any concerted effort in defining and measuring performance of network systems may lie in an ignorance of certain system performance concepts. The performance of a network system can be defined in four ways:

  Total system costs computed on a monthly basis
  System throughputs in terms of all types of transactions handled during a unit time
  Systemwide quality-of-service (QOS)
  Systemwide grade-of-service (GOS)

TOTAL MONTHLY COSTS

Transmission facilities determine the majority of the total monthly cost of MANs and WANs paid to the local exchange carrier (LEC), interexchange carriers (IECs), and other common carriers. The other major costs are for hardware and the recurring price of network management and control (NMC). Financing the hardware can turn a large one-time cost into an affordable monthly cost. The NMC costs related to spares can be handled just like one-time hardware costs. Some companies hire in-house NMC specialists; others prefer to outsource.

SYSTEM THROUGHPUT

System throughput is measured by the rate at which the various types of transactions are handled per unit time (usually second or minute). Throughput is defined by the number of call attempts or calls completed per second for a voice network. In a data network, throughput is defined by the number of packets or bits handled per second. The throughput capability of each node is generally defined by the equipment vendor. The challenge lies in measuring the system throughput. System throughput can be estimated by enumerating the exact paths of each transaction.

SYSTEM QUALITY-OF-SERVICE (QOS)

Performance aspects dealing with transmission quality, perceived voice quality, error-free seconds, data security, and network reliability (mean time between system failures) fall into the QOS criterion. Most of these parameters are very hard to compute for the entire system. Performance aspects of a critical resource can be estimated to get a feel for the quality of service of the entire system.

SYSTEM GRADE-OF-SERVICE (GOS)

The GOS criterion deals with end-to-end blocking for a voice network and average response time (measured as the elapsed time between the moment the send key is pressed and the moment the return reply is discerned by the user) for data communications. Analytical tools are available for estimating GOS parameters for voice, data, and integrated networks.

DEFINING ENTERPRISE PERFORMANCE GOALS

Performance goals for enterprise networks are generally developed by corporate strategic planners. A typical strategic planning cycle lasts several years and entails:

  Continuous evaluation of the needs of the enterprise and its competitors. This activity defines the relationship of system response times to user productivity for each transaction.
  Study of evolving new technologies, CPE, and networking standards. The most effective way of deploying these new technologies should also be investigated. This study should establish the cost and performance attributes of new hardware (e.g., ATM and LAN switches).

A network planning group should work closely with the IT department. It is better not to outsource strategic planning because an outside group cannot fully understand the close synergy between the demands of the marketplace, corporate IT, user productivity, and network operations.

Network managers today have to deal with ever-increasing demands for:

  Voice, video, image, and data communications.
  Multiplexing of digitized voice, image, and video signals with regular data traffic at all hierarchies of enterprise locations through switches (e.g., ATM switches).
  Unscheduled or varying demands for digital bandwidth at all hours of a day on a dynamic basis.

To design an integrated enterprise network, the strategic planning group needs a user-friendly tool for quickly evaluating solutions that take user demands into account. The right tool should help the strategic planning group reach solutions iteratively and interactively.

MAJOR NETWORK DESIGN ISSUES

No single approach to network design is ideally suited for all enterprises. Network design is basically concerned with two issues:

1.  Topological optimization, which determines the way network nodes are connected to one another (including the type of connections) while satisfying a set of critical design and performance constraints.
2.  System performance dealing with end-to-end response times, path congestion, and availabilities. Recurring network cost is generally the most important performance criterion and it is mainly determined by its topology. Network topology also determines the remaining performance issues such as response times and availability. Each network design package analyzes these performance issues in only an approximate manner.

PREVIOUS NETWORK DESIGN TECHNOLOGY

Many older network design tools handled only voice or multidrop data networks. Some of the tools that came later handled only interconnections of data LANs to achieve an enterprise data WAN. Furthermore, most of these tools required mainframes. The use of a mainframe introduced an unnecessary curtain between the network designer and the host processor. The network design jobs were entered invariably via the “batch” approach, and the outputs came in the form of large printouts after a good deal of delay. Each change of a design parameter or study of a new technology required a new non-interactive delay. The absence of network-related graphics from outputs caused additional delays in interpreting the significance of results.


Previous Table of Contents Next

Copyright © CRC Press LLC