Costs of a Federated and Hybrid Cloud Environment Aimed at MapReduce Video Transcoding

TitleCosts of a Federated and Hybrid Cloud Environment Aimed at MapReduce Video Transcoding
Publication TypeConference Proceedings
Year of Conference2015
AuthorsPanarello, A., M. Fazio, A. Celesti, M. Villari, and A. Puliafito
Conference Name2015 IEEE Symposium on Computers and Communication (ISCC), Larnaca Cyprus
Date Published2015
PublisherIEEE Computer Society
Conference LocationLarnaca, Cyprus
KeywordsApache Hadoop, Big Data, CLEVER, cloud computing, HDFS, Horizontal Federation, IEEE P2302, MapReduce

In this paper we investigate the applicability of the federation among several Cloud platform, demonstrating that a federated environment provides evident benefits despite the costs for the setup and maintainance of the federation itself. Also, we propose a new solution able to manage resource allocation in federated Clouds where resource requests occur in a dynamic way. We adopt such a solution to setup distributed Hadoop nodes of virtual clusters for the parallel MapReduce processing of large data sets. To increase their capabilities, Cloud Providers establish a federation relationship, making the Hadoop-based Cloud platforms much more performing than in the isolate case, adding a further level of parallelization in service provisioning. The results analyzed in the referece use case, that is a video transcoding using the MapReduce paradigm in a federated fashion, show how the federation costs in terms of delays and overhead are low in comparison with the service provisioning costs, and also highlight how federation makes the offered Cloud service more streamlined and fast.