An Adaptive Scheduling Mechanism for Elastic Grid Computing

摘要

Server utilization is typically low (10%-30%) in today's datacenters (or clouds), especially when executing computational jobs with deadlines. Previous studies have shown that it is difficult to improve utilization above 20% without significantly increasing the failure rate of job execution. It is still unknown how to increase utilization while maintaining a low (e.g., 1-5%) failure rate. To solve this problem, this paper proposes to build an elastic grid, utilizing virtual computers from clouds, which can dynamically adjust its computing capability to maximize utilization within the constraint of an expected failure rate. At the heart of the elastic grid approach is a decentralized scheduling mechanism, based on a new risk-hedged-pricing instrument. The scheduling mechanism can change the number of virtual computers in the elastic grid dynamically, on demand of the computational job workload. The risk-hedged-pricing instrument relates price to failure rate, thus can control the failure rate via automatic price adjustment. Performance evaluation is conducted via simulations, utilizing both synthetic and real workloads. The results show that our approach outperforms other schemes, improving utilization to over 90% with failure rate still less than 6.27%.