Processing Software Aging through Swarm Intelligence

Abstract

Software aging is a phenomenon that refers to
progressive performance degradation or transient failures or even crashes in long running software systems such as web servers. It mainly occurs due to the deterioration of operating system resource, fragmentation and numerical error accumulation. A primitive method
to fight against software aging is software rejuvenation. Software rejuvenation is a proactive fault management technique aimed at cleaning up the system internal state to prevent the occurrence of more severe crash failures in the future. It involves occasionally stopping the running software, cleaning its internal state and restarting it. An optimized schedule for performing the software
rejuvenation has to be derived in advance because a long running application could not be put down now and then as it may lead to losses. This paper proposes a method to derive an accurate and optimized schedule for rejuvenation for a web server (Apache) by using Multi-Layer Perceptron (MLP) feed forward Artificial Neural Network (ANN) with Particle Swarm Optimization (PSO), a swarm intelligence based algorithm as the learning algorithm. Aging indicators are obtained through experimental setup involving Apache web server and clients, which acts as input to the MLP neural network. This method is better than existing ones because usage of
PSO algorithm leads to better convergence in speed and accuracy.