Fault tolerant shared-object management system with dynamic replication control strategy

This paper is based on a dynamic replication control strategy for minimizing communications costs. In dynamic environments where the access pattern to share resources cannot be predicted statically, it is required to monitor such parameter during the whole lifetime of the system so as to adapt it to new requirements. The shared-object management system is implemented in a centralized manner in which a master processor deals with the serialization of invocations. On one hand, we attempt to provide fault tolerance as a way to adjust the system parameters to work only with a set of correct processors so as to enhance system functionality. On the other hand, we attempt to furnish availability by masking the failure of the master processor. A new master processor is elected that resumes the master processor processing. Our shared-object management system modularity is realized through a meta level implementation.