IBM Systems Journal

SRIRAM: A scalable resilient autonomic mesh

View publication


Mirroring and replication are common techniques for ensuring fault-tolerance and resiliency of client/server applications. Because such mirroring and replication procedures are not usually automated, they tend to be cumbersome. In this paper, we present an architecture in which the identification of sites for replicated servers, and the generation of replicas, are both automated. The design is based on a self-configuring mesh of computers and a communication mechanism between nodes that operates on a rooted spanning tree. A query-search component uses Java™ language-based query capsules traveling along the branches of the spanning tree, and a caching scheme whereby the query and previous search results are cached at each node for improved efficiency. Furthermore, a security and anonymity component relies on one or more authentication servers and an anonymous communication scheme using link local addresses and indirect communication between the nodes via the spanning tree. The architecture also includes components for resource advertising and for application replication.