Amalgamation of Neural Network and Genetic Algorithm for Efficient Workload Prediction in Data Center

Prediction is one among the major challenging tasks in today’s research area for predicting dynamic workload in cloud data center. Prediction is more tedious because of huge fluctuations in workloads coming to servers. Neural network is the best technology which is used in prediction using its various flavors like recurrent neural network, genetic regression neural network, etc., for prediction. Recently, bio-inspired algorithm like particle swarm algorithm, differential evolution, covariant matrix adaptation, Evolutionally algorithm, spider monkey algorithm, etc., are playing significant and consequential role in optimizing neural network for updating its weights. In this paper, we have combined neural network with genetic algorithm for better optimization of algorithm for prediction. Many researchers have ignored the resource parameters by considering only CPU as resource constraint for prediction. In this work, we have considered CPU, memory and disk storage of Google server utilization to improve the accuracy of prediction. The work is carried out using real world workload traces from Google. Comparison is done between NN and proposed hybrid load prediction algorithm (HLPA). The proposed algorithm reduces the mean square error up to 0.002 compared to neural network.