Table of Contents Author Guidelines Submit a Manuscript
Scientific Programming
Volume 2016 (2016), Article ID 7241928, 13 pages
Research Article

Feedback-Based Resource Allocation in MapReduce-Based Systems

1OEG, ETS de Ingenieros Informáticos, Universidad Politécnica de Madrid, Campus de Montegancedo, s/n Boadilla del Monte, 28660 Madrid, Spain
2Inria Rennes-Bretagne Atlantique Research Centre, Campus Universitaire de Beaulieu, Rennes, 35042 Brittany, France

Received 14 January 2016; Accepted 28 March 2016

Academic Editor: Zhihui Du

Copyright © 2016 Bunjamin Memishi et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


Containers are considered an optimized fine-grain alternative to virtual machines in cloud-based systems. Some of the approaches which have adopted the use of containers are the MapReduce frameworks. This paper makes an analysis of the use of containers in MapReduce-based systems, concluding that the resource utilization of these systems in terms of containers is suboptimal. In order to solve this, the paper describes AdaptCont, a proposal for optimizing the containers allocation in MapReduce systems. AdaptCont is based on the foundations of feedback systems. Two different selection approaches, Dynamic AdaptCont and Pool AdaptCont, are defined. Whereas Dynamic AdaptCont calculates the exact amount of resources per each container, Pool AdaptCont chooses a predefined container from a pool of available configurations. AdaptCont is evaluated for a particular case, the application master container of Hadoop YARN. As we can see in the evaluation, AdaptCont behaves much better than the default resource allocation mechanism of Hadoop YARN.