EDF Scheduling Technique for Private Cloud Environment using Hadoop MapReduce

Abstract

Job scheduling is a key issue, especially in private cloud environment where resources are limited. The importance of job scheduling increases more when an application oriented constraints such as time has to be considered, where user jobs have a deadline to meet. Two-level job scheduling scheme using EDF MapReduce framework in private cloud environment is implemented in proposed work. A first-level scheduler, the “Job Scheduler” determines the order of execution of each incoming jobs and the second level scheduler, the “Hadoop MapReduce Server” performs the actual Map and Reduce tasks. The efficiency of proposed Earliest Deadline First is illustrated over three other scheduling techniques: FIFO scheduling, Shortest Job First scheduling and Priority Based scheduling. Experimental result shows that EDF MapReduce scheduling technique leads to the lowest executing time and lowest average waiting time for job sets as compared to other three scheduling techniques mentioned above. In EDF MapReduce scheduling, almost each job in each job sets completes their execution within their respective deadlines; hence it has almost no deadline miss or sometimes at the high load time it leads to very less deadline miss as compared to FIFO scheduling, Shortest Job First scheduling and Priority Based scheduling.