Cloud computing shares data and oers services transparently among its users. With the increase in number of users of cloud the tasks to be scheduled increases. The performance of cloud depends on the task scheduling algorithms used in the scheduling components or brokering components. Scheduling of tasks on cloud computing systems is one of the research problem, Where the matching of machines and completion time of the tasks are considered. Tasks matching of machines problem is that, assume number of active hosts are Y, number of VMs in each host are Z. Maximum number of possible Virtual Machines(VMs) to schedule a single task is (y*z). If we need to schedule X tasks, number of possibilities are (y *z)^x. So scheduling of tasks is NP Hard problem. NP Hard means this scheduling of tasks on VMs not having polynomial time complexity, but it may have algorithm for verifying solution. Fault-tolerance becomes an important key to establish dependability in cloud computing system. In task scheduling, if task not completed in it's deadline ,then it is one type of fault in scheduling of tasks. In this thesis this type of faults are taken and try to overcome it. In this thesis we present a non-preemptive scheduling algorithm, By inserting the ideal time for postponing the task by ensuring the other task will completes its execution with in the deadline. In simulation the proposed algorithm maximizes the prot of 25%, throughput of 25% and minimizes the penalty of 20% over EDF.
展开▼
