Abstract

Cloud computing provides enchanting option for businesses to lease a suited size MapReduce cluster, use resources as a service, pay up only for resources that were used. A major challenge in such an environment is to enhance the consumption of MapReduce clusters to understate their cost. One of the way for obtaining this goal is to make execution of MapReduce jobs on the cluster optimum. This paper is considering MapReduce framework, Hadoop File System, the various work on scheduling in MapReduce. In addition to that, task level scheduling algorithms to address budget and deadline restraints for MapReduce workflow is considered. This has been done on heterogeneous machines in clouds. Heterogeneity is demonstrated in the “pay-as-you-go” model where the machines with varying performance would have varying service rates.

Authors and Affiliations