Report ID
2002-02
Report Authors
Kai Shen, Tao Yang, and Lingkun Chu
Report Date
Abstract
This paper studies cluster load balancing policies and system support forfine-grain network services. Load balancing on a cluster of machines hasbeen studied extensively in the literature, mainly focusing on coarse-graindistributed computation. Fine-grain services introduce additional challengesbecause system states fluctuate rapidly for those services and systemperformance is highly sensitive to various overhead. The main contributionof our work is to identify effective load balancing schemes for fine-grainservices through simulations and empirical evaluations on synthetic workloadand real traces. Another contribution is the design and implementation of aload balancing system in a Linux cluster that strikes a balance betweenacquiring enough load information and minimizing system overhead. Our studyconcludes that: 1) Random polling based load-balancing policies arewell-suited for fine-grain network services; 2) A small poll size providessufficient information for load balancing, while an excessively large pollsize may in fact degrade the performance due to polling overhead; 3)Discarding slow-responding polls can further improve system performance.
Document
2002-02.ps258.38 KB