Fuzzy Subtractive Clustering Based Prediction Approach for CPU Load Availability
Kadda Beghdad Bey, Farid Benhammadi, Faouzi Sebbak · 2013
Abstract—Distributed processing environment has emerged as a new vision for future network based calculation, allowing the federation of heterogeneous computing resources to incorporate the power. Cloud computing is a new computing paradigm composed of a combination of grid computing and utility computing concepts. In cloud computing, the prediction methods play a key role in managing large scale of computation capacity. In this paper, a modelling approach to predict the future CPU load value is presented. The proposed approach employs a computational intelligence technique to classify the CPU load time series into similarity component group. This technique is based on the Fuzzy Subtractive Clustering algorithm and a combination of local Adaptive Network-based Fuzzy Inference System. The results of an exhaustive set of experiments are reported to validate the proposed prediction model and to evaluate the accuracy of their prediction. Experimental results demonstrate both feasibility and effectiveness of our approach that achieves important improvement with respect to the existing CPU load prediction models. Keywords-Subtractive clustering; CPU load prediction; cloud computing; system modelling; ANFIS. I.