Overview

The introduction of new kinds of network applications, such as the World Wide Web (WWW), multimedia, and e-commerce, has made it increasingly difficult to accurately predict traffic behavior on large communications networks. The complex interaction between multiple protocols and applications, different vendor devices, sophisticated queuing strategies, and mixed media types all make this a complicated task.

Before investing valuable time and money in deploying a new network infrastructure, it is often worth investing some effort in trying to predict just how that network will behave. Careful design and a solid understanding of the appropriate tools to use can reduce many pitfalls associated with growth. We will examine both the general approach you should take when tackling a design and some of the underlying theory and tools you should be aware of. Discussions focus on the following topics:

• Hierarchical design models

• Topology designs

• Resilient design techniques

• Multiprotocol or single-protocol models

• Shortest path algorithms

By creating better models of network traffic we are more able to design better protocols, better network topologies, better routing and switching hardware, and better services to users. Apart from providing the predictive models it also provides an overview of key design methodologies, together with some of the key terminology used. Once the design has been decided, there are several techniques that can be employed to model network and system behavior. Generally these techniques fall into three broad categories, in increasing order of accuracy, as follows:

• Analytical models

• Simulation models

• Empirical models

Networks naturally expand to offer...