Systems exhibit a continuous degradation of performance with time. This phenomenon is well known for system administrators. Downtime is usually planned to avoid system failure. This research aims at predicting the performance degradation so as to avoid or minimize the need for bringing systems down for maintenance.

This work was conducted within IBM Center For Advanced Studies and Research (CAS). The approach deployed Neural Networks to learn patterns of system performance. The data was collected from a number of Linux servers running at a large client site. The approach was able to come to a good prediction of real life running systems. This work resulted in two publications; “Mining software aging patterns by artificial neural networks” and “Mining software aging: A neural network approach“.