Resource provisioning is one of the main challenges in resource sharing environments such as InterGrid. Recently, many resource management systems in re- source sharing environments use lease abstraction and virtual machines for provisioning. In resource sharing environments resource providers serve requests from external (grid) users along with their own local users. The problem arises when there is not sufficient resources for local users, which have higher priority than grid users, and need resources urgently. This problem could be solved by preempting leases from grid users and allocating them to the local users. However, preempting leases entails determining which lease(s) are better choices to be preempted and what should be done with the preempted leases. To answer these questions, in this work, we propose different request types in the InterGrid environment. Then, we propose and compare several policies that determine the proper set of lease(s) for preemption. The first policy increases resource utilization as a system centric criterion. The second policy improves user satisfaction by decreasing the number of preempted leases. The third policy makes a trade-off between resource utilization and the number of lease preemption. Simulation results demonstrate that the proposed preemption policies serve up to 72% more local requests without increasing the rejection ratio of grid requests.
|Cite as: Salehi, M. A., Javadi, B. and Buyya, R. (2011). Resource Provisioning based on Leases Preemption in InterGrid. In Proc. Australasian Computer Science Conference (ACSC 2011) Perth, Australia. CRPIT, 113. Mark Reynolds Eds., ACS. 25-34 |
(local if available)