Resource availability is an important issue nowadays due to the possibility of occurrences of faults in heterogeneous computing systems such as grid and cloud systems. Traditional advance reservation scheduling techniques do not consider resource availability while making resource reservations. There are many high priority jobs that need to be executed as quickly as possible but others jobs lead to the starvation of such jobs. To overcome this problem advance reservation scheduling is considered to allocate the resources to such jobs when needed which consider deadline constraint. Such jobs could not be risked to be executed on the machines that could fail very easily. The fault awareness is very important to consider which has not been done so far with advance reservation. Reliability is handled in the proposed approach by considering mean time between failures metric. Deadline sensitive jobs are compared against finish time of jobs which is calculated in advance and AHP is applied to determine their completion time. AHP matrix indicates if jobs can satisfy deadline or not. To resolve the faults, task execution state at current machine is migrated to other machine using checkpointing approach which leads to reduction of the execution time. The result of the proposed fault-aware scheduling policy showed improvement in terms Makespan and Flowtime by 10% and 14% respectively.
展开▼