Channel Islands in a Reflective Ocean: Large Scale Event Distribution in Heterogeneous Networks

• Authors: Jon Crowcroft
• Venue: NETWORKING '02 Proceedings of the Second International IFIP-TC6 Networking Conference on Networking Technologies, Services, and Protocols; Performance of Computer and Communication Networks; and Mobile and Wireless Communications
• Year: 2002

This is a discussion paper about the possible future use of network and transport level multicast services to support extremely large scale event distribution.To date, event notification services[40] have been limited in their scope due to limitations of the infrastructure At the same time, Internet network and transport layer multicast services have seen limited deployment due to lack of user demand (with the exception more recently of streaming services, e.g. on Sprint's US core network, and in the Internet II). Recent research in active and reflective middleware suggests a way to resolve these two problems at one go.Event-driven and messaging infrastructures are emerging as the most flexible and feasible solution for enabling rapid and dynamic integration of legacy and monolithic software applications into distributed systems. Event infrastructures also support deployment and evolution of traditionally difficult-to-build active systems such as large-scale collaborative environments and mobility aware architectures.Event notification is concerned with propagation of state changes in objects in the form of events. A crucial aspect of events is that they occur asynchronously. Event consumers have no control over when events are triggered. On the other hand, event suppliers do not generally know what entities might be interested in the events they provide. These two aspects clearly define event notification as a model of asynchronous and de-coupled communication, where entities communicate in order to exchange information, but do not directly control each other.The IETF is just finishing specifying a family of reliable multicast transport protocols, for most of which there are pilot implementations. Key amongst these for the purposes of this research is the exposure to end systems of router filter functionality in a programmable way, known as Generic Router Assist. This is an inherent part of the Pragmatic General Multicast service, implemented by Reuters, Tibco and Cisco in their products, although it has not been widely known or used outside of the TIBNET products until very recently.The goal of this paper is to describe a reflective middleware system that integrates the network, transport and distributed middleware services into a seamless whole.The outcome of this research will be to integrate this 'low-level' technology into an event middleware system, as a toolkit as well as evaluation of this approach for massive scale event notification, suitable for telemetry, novel mobile network services, and other as yet unforeseen applications.