Economy driven real-time scheduling for cloud
R.L. Kashyap, Paritosh Louhan, Manish Mishra · 2016
Real-time scheduling of Virtual Machines (VMs) has always been challenging for service providers, when economics of renting services on cloud need to be considered together with the deadline constraints. The diversity among VMs and price based urgent demands makes scheduling more complex when aim of service provider is to maximize the revenue. For maximization of revenue the schedule should have high success rate but at the same time preference should be given to higher paying clients. The work compares, proposed Urgency Factor Prioritized (UFP) scheduling policy, which is a greedy approach wherein urgent tasks are charged more and promoted, with Earliest Deadline First (EDF), which is yet optimal solution to maximize success rate. The observations suggest that selection and preference of the scheduling policy should depend on urgency and workload of VMs. Finally, for maximization of revenue, an economy driven real-time scheduling policy is suggested which takes best of UFP and EDF. The work implemented UFP and EDF in credit allocation strategy of Xen hypervisor. Strategies for deadline a nd cost calculations are derived considering the urgency, workload associated and real-time requirements of a VM.