Self-Adapting Credit-based Server Load Balancing

L. Schneidenbach, B. Schnor, J. Zinke, and J. Lehmann (Germany)

Keywords

Cluster Computing, Server Load Balancing, SelfAdapting Systems, Heterogeneous Systems

Abstract

Clusters are a popular platform to build highly available and scalable service solutions. Usually, the front-end dis patcher distributes the load to the back-end servers in a round-robin fashion or according to current server load based on external load parameters. The use of server weights is very popular to improve the load balancing in case of heterogeneous server machines. Correctly deter mined weights are crucial to the quality of the distribution. While the determination of weights can be done in small and static environments, it can hardly be done in dynamic or heterogeneous environments. In this paper, we present a credit-based algorithm for server load balancing with a complexity of O(1) that adapts to heterogeneous environments without the need to specify server weights. This approach is able to self-adapt to het erogeneous servers and heterogeneous workloads. Cred its are calculated by using available communication end points. We present simulation results as well as experimen tal results with our prototype where the update of credit in formation is done efficiently using the RDMA capabilities of InfiniBand.

Important Links:



Go Back