An automatic network troubleshooter for throughput bottlenecks in computer networks

Government agencies operate several large computer networks; diagnosing and repairing performance problems on these networks is an expensive task. The Automatic Network Troubleshooter (ANT) project's goal is to reduce this expense by reducing the mean-time-to-repair for performance problems. ANT uses a model-based troubleshooting technique: first creating an executable model of network misbehavior, then exploring it to identify the cause of performance problems. ANT uses pattern matching and the protocol hierarchy to speed up the search for the bottleneck. Finally ANT uses sensitivity analysis to determine the appropriate corrective action. The ANT prototype described detects throughput bottlenecks caused by misconfigured parameters and lack of bandwidth.<
>