A Novel Adaptive Distributed Load Balancing Strategy for Cluster

* This paper is supported by National 863 Hi-Tech R&D Project under grant No.2002AA1Z2102 Abstract with the dramatically increasing of users accessing Internet services, many researchers employed cluster server to provide their Internet services and proposed all kinds of scheduling strategies to address the imbalance among real servers. However, most of them are based on TCP connections and pay little attention to the diversity of requests and real servers. Thus these strategies will lead to serious imbalance if deployed in a heterogeneous cluster. In this paper, we propose a novel adaptive distributed load balancing strategy, called ADLB, to address this problem. Different from the existing solutions, it offers a kind of least delay scheduling method, which implements adaptive scheduling strategy by the TCP-SYN packet delay mechanism and doesn’t require any additional communication among real servers and front-end dispatcher participation. Experimental results show that our technologies can achieve better balance for heterogeneous cluster and provide good scalability.