Genetic Algorithm and PID Control Together for Dynamic Anticipative Marginal Buffer Management: An Effective Approach to Enhance Dependability and Performance for Distributed Mobile Object-Based Real-Time Computing over the Internet

Abstract

A novel model that combines genetic algorithm (GA) and proportional + integral + derivative (PID) control for adaptive marginal buffer management is proposed here. The goal is to prevent buffer overflow at the receiver side so that message retransmissions that lead to poor system reliability and performance can be eliminated. Marginal buffer control keeps the difference between the buffer length and the queue length continuously within a safety margin Δ, in an anticipative, adaptive manner. At first, we will propose the PID controller and show that it alone can achieve the goal but with some shortcomings. We then propose to trim these shortcomings by the GA objective function {0, Δ} 2. The PID control makes use of the micro version of the convergence algorithm, namely, the M 2RT, which is an effective internet end-to-end performance measurement ( IEPM) method developed by us previously to predict the trend of a distribution. In this case the M 2RT, which exists as an independent program object, predicts the dynamic queue length quickly and accurately once invoked by clients. Such predictions enable the GA-augmented PID (GA–PID) to determine whether the buffer length should be increased or decreased adaptively to maintain the safety margin Δ. The GA–PID model was verified and validated in a distributed mobile object-based real-time computing (DMORC) environment, which was implemented over the Internet with the Java-based Aglets mobile agent platform. The preliminary tests confirm that the GA–PID model is indeed an effective solution for achieving dynamic marginal buffer management for DMORC systems.