MapReduce Model: A Paradigm for Large Data Processing

Abstract

MapReduce is a programming paradigm that enables massive processing of large amount of data over several<br> machines in a cluster of commodity computers. It is fault tolerant and scalable hence suitable for cloud computing<br> applications. This paper come up with a second order nonlinear model of the MapReduce using experimental data<br> collected from Grid5000 experimental tested accessed from the local machine using Linux Secure Socket Shell<br> protocol (SSH) as a command line interphase. System identification was performed on the collected data using<br> MATLAB toolbox. The nonlinear model obtained was linearized and discretized using numeric optimization<br> techniques to obtain the continuous transfer function. Within the limits of operating points, the model shows a<br> perfect tracking and good representation of the dynamics of the real system and hence can be suitable for applying<br> control laws.<br>