An extensible architecture for monitoring and modeling network systems.

Abstract

This dissertation presents an architecture for monitoring and modeling network systems: a client/server protocol for controlling and monitoring data collection and Management Information Bases (MIBs), designed specifically to support modeling. A prototype system based on this architecture demonstrates its ability to enhance network capacity planning, configuration, and troubleshooting. Prototype implementations are applied to real network problems and the results are analyzed in detail. The network problems used in our examples cannot be solved using current monitoring and modeling technologies; however, we find that network management is improved by providing additional methods not currently available to analyze network troubleshooting problems. Advances in network monitoring protocols, such as SNMP and RMON, combined with their widespread availability from commercial vendors, have led to the development of tools that enable system administrators to effectively manage problem domains, such as network configuration and troubleshooting. The abundance of monitoring and data-gathering tools presents an opportunity to combine network monitoring and modeling technologies to provide enhanced network management capabilities. For example, a network administrator can monitor an existing internetwork and use this information as a basis for capacity-planning projections. Alternative configurations can be compared using "live" network data to characterize the traffic streams, which is superior to using "back-of-the-envelope" estimates. We investigate the feasibility of combining monitoring and modeling technologies, as well as uncovering the difficulties of this combination, by implementing and applying a prototype architecture to a configuration/capacity planning problem. We implement an initial network configuration, collect monitored data on the initial configuration, and then show the change in performance due to different configurations by modeling changes to the initial configuration. We then implement the configuration changes to the network, measure performance results, and compare them to our modeled projected results. The monitored and projected results are shown to be quite close. Our proposed architecture is designed to use commonly available resources and can be tailored for use with many different tools for monitoring, modeling, and data storage, making our approach economically feasible. We also show how monitored data can be made available remotely, possibly to other network administrators, for use in joint problem solving.