Amrita International Conference of Women in Computing - 2013 |
Foundation of Computer Science USA |
AICWIC - Number 3 |
January 2013 |
Authors: Ashutosh Rajan, M. V. Judy |
1e38806d-8fe2-480b-b67f-d3cf849a69b7 |
Ashutosh Rajan, M. V. Judy . An Enhanced Map Reduce Framework for Improving the Performance of Massively Scalable Private Clouds. Amrita International Conference of Women in Computing - 2013. AICWIC, 3 (January 2013), 24-26.
Cloud Computing systems provide access to large amount of data and other resources through a large number of interfaces. Apache Hadoop is a framework that allows distributed processing of large sets of data across cluster of computers. It is a powerful abstraction proposed for making scalable and fault tolerant applications. In this paper we have suggested an enhanced framework for MapReduce which increased the performance of the Private Clouds in distributed environment. In this framework a separate thread is maintained for each and every Mapper and a single buffer is used for retrieving all threads. A single Buffer retrieves all records. At this instance a separate thread can search for all the records with same key in the buffer and pass it on to the Reduce function which can executed. A multimap is used to access partial result while maintain key ordering. Our analysis shows that better performance can be achieved using this enhanced Map Reduce framework than using the traditional MapReduce framework. The results show the reduction in job completion time when compared with existing one.