High available software architecture based on cluster technology

This paper presents an approach to provide 24-hour nonstop financial services using high availability software architecture based on clustering technology. The classical cluster architectures (such as shared disk, shared memory, and shared I/O bus) cannot satisfy the application requirements. The approach makes use of RPC (remote procedure call) to implement the fail-over and data recovery. Different fail-over policies are applied to different kinds of failures (such as hardware failure, process failure and so on). A prototype system has been developed and tested.