Decentralized Adaptive Routing for Reliability in Event Broker Networks

Guaranteeing quality of service (QoS) for event delivery has been recognized as an important but challenging issue in event based middleware (EBM), that is responsible for routing events from publishers to subscribers over an event broker network. Amongst the numerous QoS parameters, in our work, we focus on reliability as a service guarantee to subscribers in an EBM. We add to the existing body of work in this area by investigating reliability needs of type-specific subscriptions and comparing them with type-agnostic subscriptions. The broker network establishes routes by maintaining event-type specific path-quality information at every broker node occurring in the route to the destination. Each broker node measures the drop probability of a particular event type. We prove that the drop probabilities experienced by individual event types are proportional to the ratio of their inter-arrival times at the broker. Based on this, we present the TSAR (Type Specific Adaptive Reliability) algorithm. where route establishment is done in an adaptive and decentralized fashion using persistent type-specific path quality information stored as a matrix of reliability estimates. Our results show that TSAR (1) reduces the overall message complexity as compared to existing efforts in this area (2) provides subscribers with a higher level of granularity when subscribing to events and adapts to the dynamics of the broker network with varying reliabilities of broker nodes.