Deadline and Cost based MapReduce Job Scheduling in Heterogeneous Cloud using Dynamic Pricing

Cloud Computing has emerged as an eminent technology in internet world. Job scheduling is one of the most challenging issues in the cloud computing area. This paper is centered on scheduling algorithms for MapReduce jobs with respect to deadline and cost constraints on a range of provisioned heterogeneous machines in the cloud. Each job is partitioned into multiple tasks. In addition, the time for executing a task on clouds resources and price for using that resource is determined. We propose a dynamic pricing algorithm for dynamically changing the price required for task execution. It helps the cloud service provider(CSP) to recover its machine’s cost and various other cost required for task execution based on the scenario when job is submitted. This algorithm will benefit CSP to maximizes its revenues. In addition to that, it provides inexpensive services to users as the cost of running jobs in the cloud is reduced by dynamic pricing. The Budget Distribution algorithm takes input the dynamic time-price table and distributes total budget among all the tasks, and remaining budget to slowest task to increase overall execution time. In this approach, we use Multiple Choice Knapsack (MCKS) deadline constrained algorithm that takes the dynamic pricing table as input and optimal cost efficient machine based on execution time required on that machine. An evaluation has been carried out which verifies the efficiency of proposed algorithm and also illustrates that the technique is cost-effective to both CSP and user.