Modelling of Map Reduce Performance for Work Control and Supply Provisioning

Abstract

Data intensive applications adopts Map Reduce as a major computing model. Hadoop, an open source implementation of MapReduce, has been implemented by progressively increasing user community. Many Cloud computing service providers offer the chances for Hadoop operators to contract a certain amount of supply�s and remunerate for their usage. Nevertheless, a key contest is that cloud service providers do not have a supply provisioning mechanism to fulfil user works with target requirements. At present, it is solely the user's accountability to evaluate the required amount of supply forming a work in the cloud. This paper presents a Hadoop presentation model that precisely guesses completion time and further provisions the required amount of supply�s for a work to be completed within a deadline. The proposed model forms on past work execution records and services Locally Weighted Linear Regression (LWLR) technique to estimate the execution time of a work. Moreover, it pays Lagrange Multipliers technique for supply provisioning to satisfy works with deadline requirements. The proposed method is primarily assessed on an in-house Hadoop cluster and then evaluated in the Amazon EC2 Cloud. Experimental results show that the accuracy of the proposed method in work execution approximation is in the range of 90.37% and 91.28%, and works are completed within the required limits following on the supply provisioning scheme of the proposed model.