Event broker networks -scalable versions of the publish-subscribe paradigmact as peer-to-peer overlays on broker nodes. Various frameworks support different overlay topologies and routing schemes for event dissemination, but attention is now turning to the nonfunctional attributes (such as quality of service) of such systems. Although many research efforts are starting to address the need for adaptation and QoS, no taxonomies or comprehensive surveys of adaptive middleware, which provide support for service guarantees, exist yet. To tackle this knowledge gap, the authors examine existing event-based middleware efforts, focusing on quality of service and adaptation.
Middleware for event broker networks (EBNs) both alleviates the issues related to underlying platform heterogeneity and provides a uniform application interface. A common service interface that such middleware provides is publish-subscribe, 1 a paradigm in which producers publish information and consumers subscribe to it. In EBNs, information of interest is encapsulated in events; on such networks, middleware stores and manages subscriptions as well as routes events. This model's fully decoupled nature in terms of time, space, and synchronization 1 makes it highly suitable for large-scale distributed applications. Middleware for event broker networks is also known as eventdissemination middleware (EDM).To disseminate events, an EBN overlay network 2 uses a subset of the underlying existing physical network (an overlay link itself is virtual but could consist of several physical links to the underlying network). Therefore, middleware must execute on each overlay node that has to work cooperatively to accomplish its functions. Such event-based architectures appear in various domains, many of which themselves require quality-of-service (QoS) guarantees from the underlying infrastructure. However, our survey of existing efforts in research-and-industrybased event dissemination middleware indicates significant gaps in efforts to address these needs; in fact, we found that few middleware options 3-5 provide support for nonfunctional service guarantees. Because a comprehensive survey of current options doesn't yet exist, this article offers a review of existing event-based systems, classified with a taxonomy of adaptive, event-based middleware that provides QoS guarantees.
Middleware ArchitectureAs Figure 1 shows, event-based middleware has a layered architecture. The core comprises mandatory functional features, such as the event model, the subscription scheme used to disseminate events, and the overlay routing substrate. The layer on top of the core forms a set of optional services -that is, the nonfunctional QoS guarantees the middleware provides, such as reliability, delivery semantics, message ordering, security, and fault tolerance. QoS guarantees are orthogonal to middleware-supported core functions; the middleware's ability to reconfigure itself to ensure and maintain a QoS agreement is its adaptability, which can result in changes to the core's subs...