Connection Migration Mechanisms in a Cluster Environment

Abstract

Nowadays, grid applications and Web applications introduce significant high availability and scalability challenges to cluster database servers. In this paper, we describe a connection failover mechanism toward building a highly availably infrastructure that is resistant to partial failure. While available solutions concentrate on client transparency, in our work we achieve server transparency as well. We demonstrate that our solution is transparent to clients and servers without compromising the system efficiency. Our mechanisms do involve neither kernel changes nor changes to the server software