The visual interface for the Netvis Analyzer showing the top levels of the Internet. The panel has controls for navigation, selection, and launching the Binder and the Grapher.
The NSFNET backbone topology selected and isolated.
The Grapher for plotting data selected via the visual interface. A desktop metaphor is used with plots on transparencies that can be grouped and overlaid.
Present tools for network analysis and management are mostly for real-time troubleshooting (seeing what's up or down at a given moment) and have little capability for analyzing all the many factors that simultaneously and dynamically contribute to the performance of modern high bandwidth, highly interconnected networks. To overcome the inadequacy of these tools, we are engaged in a two-part program. In one part, we are developing methods to take network data, which is likely to be massive, multivariate, and dynamic, and investigate it effectively and efficiently. These methods employ approaches from both interactive, 3-dimensional graphics and statistics. To do this we are developing a set of tools called the Netvis Analyzer. In the other part of program, we are working with network managers and network simulation researchers to develop a set of observations and techniques that will provide the network investigator with enough data to make informed management decisions about the current state of the network and predictions about the future state of the network as it evolves in connectivity, bandwidth, and type and amount of traffic. This observational effort will be directed toward the characteristics of the NSFNET backbone, the connected regional networks (such as SURAnet), and also local networks. In this joint effort of observation and analysis, we will be searching for the most important factors affecting network stability and efficiency under various loads and types of traffic. We will then incorporate methods for easy measurement and analysis of these factors in our Netvis Analyzer tools. Ultimately we will have an integrated observation and analysis toolkit that will be placed in a programmerless, highly interactive environment.