Analytical decisional model for latency aware publish/subscribe systems on MANET

This paper proposes an analytical model for latency aware publish/subscribe systems on mobile ad hoc networks. The proposed approach combines both proactive and reactive statistical analysis. On the one hand, the reactive analysis, suitable for multimedia applications, detects failures by approximating latency series with the Gumbel distribution. On the other hand, the proactive analysis, suitable for crisis management applications, forecasts failures occurrence relying on Auto Regression or Auto Regressive Integrated Moving Average Formulas. Finally, a hybrid analysis was proposed by dynamically switching from reactive to predictive forms of analysis whenever quality of service violations are noticed. In order to extract failure cause, we refer to the correlation method once failure was detected or predicted. Simulations done under different scenarios proved the efficiency and accuracy of the proposed scheme.