An Improved Scheduling Technique based on Containers

In hadoop, job scheduling is an independent module so that users can design or configure their own job scheduler based on their actual application requirements; thereby meet their specific business needs. Currently, hadoop has three schedulers: first is FIFO, Fair scheduler and Capacity Scheduler, all the schedulers can helps in scheduling the resources among different computers. All the three schedulers cannot fully support data locality due to which the performance is affected. In this paper, we took the concept of resource-prefetching into consideration, and proposed a job scheduling algorithm based on data locality. By which the PRISM scheduler will perform much better because along with data locality we also focus on containers by which we can schedule the containers among different map-reduce task.