Real-Time Network Management

The objective of our effort was to develop methods for diagnosing network performance problems in real time. According to our Phase I research, it is possible to collect data on the network and morph it into queuing models to produce information about the network and physical layers of nodes on a network. Through our methods it is possible to collect data on the network and morph it into queuing models to produce information about the network and physical layers of nodes on a network. For Phase I, we focused on the MM1 for our implementation. Other models defined in this Final Technical Report include MM1k, MD1, and MG1. We also defined model selection rules that will allow our application to automatically select which model should be used pending the amount of utilization on the link or node. The preliminary results of this three-phase project are twofold. First we have proven that queuing models can be used to mirror the characteristics of links and nodes. Secondly, we have shown that the outputs from the queuing models can be used as input to a diagnostic subsystem, which in turn can provide the network manager with information to proactively identify possible problems.