A Dynamic Self-adaptive Resource-Load Evaluation Method in Cloud Computing

Abstract

Cloud resource and its load have dynamic characteristics. To address this challenge, a dynamic self-adaptive evaluation method (termed SDWM) is proposed in this paper. SDWM uses some dynamic evaluation indicators to evaluate resource state more accurately. And it divides the resource load into three states -- $Overload$, $Normal$ and $Idle$ by the self-adaptive threshold. Then it migrates overload resources to balance load, and releases idle resources whose idle times exceed a threshold to save energy, which can effectively improve system utilization. Experimental results demonstrate SDWM has better adaptability than other similar methods when resources dynamically join or exit. This shows the positive effect of the dynamic self-adaptive threshold.