Management of Temporally and Spatially Correlated Failures in Federated Message Oriented Middleware for Resilient and QoS-Aware Messaging Services.

Abstract

Message Oriented Middleware (MOM) is widely recognized as a promising solution for the communications between heterogeneous distributed systems. Because the resilience and quality-of-service of the messaging substrate plays a critical role in the overall system performance, the evolution of these distributed systems has introduced new requirements for MOM, such as inter domain federation, resilience and QoS support. This thesis focuses on a management frame work that enhances the Resilience and QoS-awareness of MOM, called RQMOM, for federated enterprise systems. A common hierarchical MOM architecture for the federated messaging service is assumed. Each bottom level local domain comprises a cluster of neighbouring brokers that carry a local messaging service, and inter domain messaging are routed through the gateway brokers of the different local domains over the top level federated overlay. Some challenges and solutions for the intra and inter domain messaging are researched. In local domain messaging the common cause of performance degradation is often the fluctuation of workloads which might result in surge of total workload on a broker and overload its processing capacity, since a local domain is often within a well connected network. Against performance degradation, a combination of novel proactive risk-aware workload allocation, which exploits the co-variation between workloads, in addition to existing reactive load balancing is designed and evaluated. In federated inter domain messaging an overlay network of federated gateway brokers distributed in separated geographical locations, on top of the heterogeneous physical network is considered. Geographical correlated failures are threats to cause major interruptions and damages to such systems. To mitigate this rarely addressed challenge, a novel geographical location aware route selection algorithm to support uninterrupted messaging is introduced. It is used with existing overlay routing mechanisms, to maintain routes and hence provide more resilient messaging against geographical correlated failures.