Autonomic performance management of cloud server based on adaptive control method

Abstract

Cloud computing provides a promising approach for efficiently managing the performance of servers via advanced resource management, hence it becomes one of the important hotspots in high-performance computing field recently. For the existing performance management solutions of cloud servers, they always show inefficiency issues when dealing with the dynamic and burst web workloads. In this paper, we propose an autonomic performance management of cloud servers, which adopt the linear quadratic Gaussian with stochastic method (LQGwS). In the face of dynamic and burst web workloads, it guarantees the workload balance between different Web applications by adaptively adjusting the amount of resource allocation to each virtual machine. Furthermore, in order to deal with the unknown disturbances in the Web system, the LQGwS describes the Web system as a coupled multiple-input-multiple-output system and uses the Autoregressive moving-average model with exogenous inputs model (ARMAX) firstly, and then constructs the optimal resource allocation scheme based on minimizing an average cost function among a set of models, which are generated according to a Gauss distribution. Through the test of real network load, the results of this experiment on the XEN-based platform show that the proposed control strategy has better performance than existing solutions under dynamical workloads in terms of control accuracy and stability.