GEO DISTRIBUTED PARALLELIZATION PACTS IN MAP REDUCE FRAMEWORK IN MULTIPLE DATACENTER

C.Kirubanantham, C.Rajavenkateswaran

Abstract: MapReduce framework in Hadoop plays an important role in handling and processing big data. Hadoop is scalable that is it can reliably store and process petabytes. MapReduce works by dividing input files into chunks and processing these in a series of parallelizable steps. MapReduce framework offers a response to the problem by distributing computations among large sets of nodes. we concentrate on geographical distribution of data for sequential execution of MapReduce jobs to optimize the execution time. The fixed execution strategy of MapReduce program is not optimal for many task and as it does not know about the behavior of the functions. Thus, to overcome these issues, we are enhancing our proposed work with parallelization contracts. The parallelization contracts include input and output contract which includes the constraints and functions of data execution. These contracts help to capture a reasonable amount of semantics for executing any type of task with reduced time consumption

Keywords: Bigdata, Datacenter, Geo-distributed, Hadoop, MapReduce, PACT

DOI: https://doi.org/10.15623/ijret.2014.0319030