Real Time Task Scheduling in Cloud Computing Environment

Abstract

Scheduling of real time tasks on cloud is one of the research problem, Where the matching of machines and completion time of the tasks are considered. Real time task's matching of machines problem is that, assume number of active hosts are p, number of VMs in each host are q. Maximum number of possible VMs to schedule a single task is (p*q). If we need to schedule r tasks, number of possibilities are (p*q)^r . So scheduling of tasks is NP Hard problem. Completion time constraint of real time task is that if task complete in dead line then only it is useful else it is not. If it is not useful then it is rejected. Earliest Dead line First(EDF) algorithm is well known algorithm for scheduling of real time tasks. EDF is Event Driven scheduling algorithm with priority assign as dynamically with respect to their deadlines. Real time tasks can be periodic, Sporadic, Aperiodic tasks. We have used Aperiodic and Periodic model to evaluate performance of varies scheduling algorithms. In general EDF Scheduler schedule the tasks such that it assign the task to the free available machine without considering the task on that machine will meet the dead line or not. In this work checked the completion time of task on the free available machines before assigning the task to the machine. To assign the task i have used three dierent techniques. First Fit, Best Fit, Worst Fit. Here Fit of task means that the task will complete it's execution on that machine in it's dead line time. These three techniques and Basic EDF are used in scheduling of aperiodic tasks and also periodic tasks. We have study the perfomance of the techniques First Fit EDF(FFE), Best Fit EDF(BFE), Worst Fit EDF(WFE). The simulation has carried out in house simulator using matlab by taking performance parameters as Guarantee Ratio(GT), VM Utilization (VU), and Through Put(TP). In simulation results it is shown that FFE, BFE and WFE algorithms are better in performance than the Basic EDF algorithm.