Managing network capacity based upon simulation of failures of network devices

Abstract

A system is described that combines outputs of network simulations, which are used for capacity planning, with empirical data of network element failure probabilities, to produce what traffic utilization is needed for each network device interface to achieve a specified availability target. Utilization of each interface can be captured for each failure case simulated. From this failure data, the utilization distribution for each link can be produced, across all failure cases. This utilization-by-failure data from simulation is combined with the empirical mean-time between failure (MTBF) and mean-time to recovery (MTTR) of each failure from the span reliability data set. This allows for a prediction of what percentage of time the interface is utilized. Thus, a relationship is established between the capacity provisioned and the percentage of time that the capacity will be uncongested. Network devices can then be added to the network to establish the availability target.