摘要

Large scale distributed system requires stringent quality of service(QoS), including availability and dependability. In this paper, we present a generic framework which provides a generic load balancing service environment for all kind of applications, and can improve the system availability remarkably with the experimental results. Firstly, this paper introduces the general adaptive load-balancing model(GALB/M), and its key technologies, including HA-RMI management, failover and dynamic migration. Secondly, it presented a Java-based prototype system implementation of the proposed GALB/M model. At last, it describes the experimental results of this framework using one RMI-based application as the testing data. The experiments are designed to evaluate the performance of adaptive and non-adaptive load balancing policies with various sizes of cluster.

全文