Clustering is a technique used to categorize instances of data that fall into some natural grouping. These groups form clusters of data, where all instances of a given cluster bear some strong resemblance to each other in comparison to instances of other clusters. Network managers often must analyze huge quantities of data of various sorts gathered by network equipment. Clustering can be used to sort the information and infer rules, find causes and effects, detect patterns, etc. to be able to summarize this massive amount of information.

This project investigates the use of clustering algorithms as a solution for improving the efficiency of a computer network. Network managers should be able to view the communication behaviour and patterns in the network. From this information, the network managers can see how computers collaborate in the network to achieve a specific task and decide, based on the communication behaviours, whether the topology is optimal for the communication pattern.

There are many applications that can be derived from this project. Automatic reconfiguration of a network based on its usage is an example. One possibility is to include communication pattern information in the computation of the spanning tree in order to identify the most efficient routes. Automated reconfiguration capabilities may also be used to mitigate the effect of some denial of service attacks.