Adaptive scheduling for parallel tasks with QoS satisfaction for hybrid cloud environments
The Journal of Supercomputing
| Hi-index | 0.00 |
Task scheduling problems are of paramount importance which relate to the efficiency of the whole cloud computing facilities. In Hadoop, the open-source implementation of MapReduce, scheduling policies, such as FIFO or delay scheduling in FAIR scheduler is used by the master node to distribute waiting tasks to computing nodes (slaves) in response to the status messages of these nodes it receives. Although delay scheduling policy has claimed to improve the throughput and response times by a factor of 2 compared to FIFO policy, it can still achieve more improvement by considering a holistic view of all the tasks waiting to be processed. Therefore, this paper proposes a new scheduler which makes a scheduling decision by evaluating the entire group of tasks in the job queue. A genetic algorithm is designed as the optimization method for the new scheduler. The preliminary simulation results show that our scheduler can get a shorter make span for jobs than FIFO and delay scheduling policies and achieve a better balanced load across all the nodes in the cloud.