Clustering of CPU Usage Data in Grid Environment using Evoc Algorithm

Abstract

Clustering is a process of organizing objects into groups whose members are similar in some way. It is one of the data mining techniques is an unsupervised learning. In a grid environment, the number of computing nodes and users may reach up to thousands or millions. The grid is said to be dynamic in that the behaviors and values of these resources change all the time. Hence, these data are not suitable to be processed in an off-line mode. The existing clustering techniques today however emphasize more on the data’s behaviors categorization but not the data’s stability. Furthermore, the normal clustering techniques are more suitable to be used for static data type in an off-line mode. This paper addresses these issues by presenting an Evolving Clustering (Evoc Algorithm) which is an improved version of Evolving Clustering Method (ECM). We apply both methods on CPU usage to identify computers behaviors. The algorithm has been evaluated using three main criteria; that is dynamicity, accuracy and the ability to identify the stable cluster members. Our results show the improvements of the algorithm to process the data in an on-line mode in the evaluation of the algorithm’s dynamicity and accuracy criteria compare to other existing clustering techniques. Furthermore, the stability evaluation was a success where we were able to identify the stable cluster members from the filtered stable clusters. However, the result was highly affected by three factors namely threshold value, stability value and stability hour.