Resource-utilization-aware task scheduling in cloud platform using three-way clustering

Abstract

Task clustering is an effective approach of improving cloud computing resource utilization, which includes other benefits such as better QoS, load balance and low energy consumption. Different existing clustering methods have sharp boundaries, three-way clustering as an application of three-way decision, uses core region and fringe region to represent a cluster. In this paper, we propose a novel idea of clustering weight algorithm called TWCW algorithm(Three-way clustering weight) based on three-way decision to overcome the low utilization aiming at improving energy-efficient. The algorithm encompasses two steps, the identified tasks are assigned into the core region and the uncertain tasks are assigned into the fringe region based on diversity of cloud tasks and the dynamic nature of resources using the three-way K-means clustering firstly. The cluster center of CS _i , centroid _i  = {mips, ram, bw} is obtained from the result of three-way clustering. In the second step is to score clusters and schedule tasks. We define a scoring matrix to record scores of the weight between clusters and the preference of attributes within clusters according to the cluster center, and then schedule tasks based on scoring matrix. We validate the high utilization of resources of the proposed algorithm by using simulation of CloudSim. The experiment shows the proposed algorithms significantly reduce energy consumption while significant improving response time of tasks comparing with K-means algorithm and FCM algorithm.