Fault-Tolerant Cluster Management Tool for Supporting Nonstop of Cluster DBMS

摘要

In this paper, we design and implement a fault-tolerant cluster management tool which can monitor the status of the nodes in a cluster DBMS (Database Management System) as well as the status of the various DBMS instances at a given node. Based on the monitoring, our cluster management tool can provide an intelligent fail-over mechanism for dealing with multiple node failures in a cluster DBMS, thus making its fault-tolerance possible when combined with a load balancer. We implement our cluster management tool by using Linux virtual server 0.81 as a load balancer and iBASE/Cluster as a cluster-based DBMS, under a cluster system with four server nodes. It is shown that the sensing and recovery time of our cluster management tool for the iBASE/Cluster is about 2 seconds while that of OCMS for Oracle DBMS is more than 20 seconds. Finally we show that our cluster management tool can support the nonstop and self-managing of cluster DBMS with its intelligent fail-over mechanism even in case of multiple node failures.