An example of automatically troubleshooting a throughput bottleneck using model based techniques

A detailed example is given for the process of troubleshooting a throughput bottleneck, and the techniques used by human analysts are contrasted with those used by an automated tool called ANT (automatic network troubleshooter). Troubleshooting a throughput bottleneck involves finding specific parameters out of hundreds of possible parameters that can be tuned to increase throughput. The automated tool uses an expert system to speed the search by reducing the complexity of a performance model. Sensitivity analysis is done on the simplified model to identify parameters that can be changed to increase throughput. The output of the tool is a simple model of the bottleneck process and a list of parameters that can be tuned. ANT uses shallow reasoning expert systems to give hints about possible locations for the bottleneck, but backs the suspicions with deep reasoning about the underlying mechanisms used by the network. A prototype tool demonstrates that combining performance modeling with expert systems allows for faster and more accurate troubleshooting that can be done by a human analysis