Research on Scheduling Algorithms in Web Cluster Servers

This paper analyzes quantitativelythe impact of the load balance scheduling algorithms and the localityscheduling algorithms on the performance of Web cluster servers, andbrings forward the Adaptive_LARD algorithm. Compared with therepresentative LARD algorithm, the advantages of the Adaptive_LARD arethat: (1) it adjusts load distribution among the back-ends through theidea of load balancing to avoid learning steps in the LARD algorithmand reinforce its adaptability; (2) by distinguishing between TCPconnections accessing disks and those accessing cache memory, it canestimate the impact of different connections on the back-ends' loadmore precisely. Performance evaluations suggest that the proposed methodoutperforms the LARD algorithm by up to 14.7%.