Failure Prediction for Publish/Subscribe System on MANET

In this paper, we treat the quality of service challenge for publish/subscribe system deployed in MANET. The proposed approach provides monitoring and proactive analysis at the middleware layer for predicting QoS degradations of logical links between brokers. Failure prediction is performed using a statistical method called Auto Regression (AR) which allows to analyze the monitored data in order to avoid failures occurrence. The simulation results which prove the efficiency of the proposed approach are reported.