QoS-aware connection resilience for network-aware grid computing fault tolerance

Current grid computing fault tolerance leverages IP dynamic rerouting and schemes implemented in the application or in the middleware to overcome both software and hardware failures. Despite the flexibility of current grid computing fault tolerant schemes in recovering inter-service connectivity from an almost comprehensive set of failures, they might not be able to repristinate also connection QoS guarantees, such as minimum bandwidth and maximum delay. This phenomenon is exacerbated when, as in global grid computing, the grid computing sites are not connected by dedicated network resources but share the same network infrastructure with other Internet services. This paper aims at showing the advantages of integrating grid computing fault tolerance schemes with next generation networks (NGNs) resilient schemes. Indeed, by combining the utilization of generalized multi-protocol label switching (GMPLS) resilient schemes, such as path restoration, and application or middleware layer fault tolerant schemes, such as service migration or replication, it is possible to guarantee the necessary QoS to the connections between grid computing sites while limiting the required network and computational resources.