Date of Publication :7th September 2017

Abstract: Now a day’s size of the data used in today’s enterprises worlds has been growing at exponential rates day by day. This had triggered need to process and analyze the large volumes of data for business decision making quickly as well. MapReduce is considered as a core-processing engine of Hadoop, which is prominently used to cater continuously increasing demands on computing resources imposed by massive data sets. Highly scalable feature of MapReduce processing, allows parallel and distributed processing on multiple computing nodes. This paper talks about various scheduling methodologies and most appropriate one can be used for improving MapReduce processing .Also tried to identify scheduling methods scaling or processing limitations along with the situations wherein they can be best suited. Map Reduce is used majorly for short jobs, which eventually require low response time. The current Hadoop implementation assumes underline computing nodes in a cluster are homogeneous, have same processing capability and memory. Hadoop’s scheduler suffers from severe performance degradation in heterogeneous environments. In heterogeneous environment, Longest Approximate Time to End (LATE) scheduling can be most efficient in comparison to other scheduling .It has been seen in various studies that LATE has improved Hadoop response times by approximately two times in a clusters.

Reference :