An adaptive approach for ensuring reliability in event based middleware

Abstract

In this effort we tackle the important issue of providing Quality of Service (QoS) guarantees to subscribers, when routing event notifications in a publish-subscribe domain. In particular, we focus on providing reliability as a QoS guarantee, which is a proportion of published events received by a subscriber. We realize this by setting up routes that guarantee subscriber requested reliability on an event broker network where every broker node is associated with a dynamically changing node reliability value. Route determination is done in an adaptive fashion via our AR (Adaptive Reliability) algorithm that uses persistent path-quality information, in the form of reliability-estimates and (1) minimizes number of messages transmitted during route-establishment (2) guarantees reliable event notification delivery to the subscribers and (3) subsequently uses reliability-estimates to refine route quality adaptively. We validate the efficacy of AR empirically using DR-SIM (Dynamic Reliability Simulator), that we have built over Hermes. DR-SIM measures dynamically changing reliability values of every broker node in the network. Our results reveal that AR guarantees reliable event notifications to subscribers, with a (1) lower message complexity compared to other existing efforts in this area, (2) is scalable in terms of space consumed with increasing sizes of broker networks and increasing number of clients (3) and is able to adapt to the dynamics of varying node reliabilities in the broker network.