ABSTRACT -- High-availability support for Beowulf cluster becomes a critical factor in the acceptance of this platform for mission critical use in enterprise environment. A well defined and extensible HA software architecture is needed. This paper presents the proposed high-availability software architecture called AMATA. AMATA architecture clearly defines the software component and interaction for High Availability support. Many cases such as system software failure, registered user software error, hardware mal-function, and hardware overload can be detected and handle in a systematic way. Both discovery and recovery process can be added to provide an intelligent and automatic fault recovery process. Currently, a prototype implementation has been developed and the results obtained from that implementation have also been included.

KEYWORDS -- High availability, Beowulf clusters, Fault Tolerance