A hybrid multicast-unicast infrastructure for efficient publish-subscribe in enterprise networks

Abstract

One of the main challenges in building a large scale publish subcribe infrastructure in an enterprise network, is to provide the subscribers with the required information, while minimizing the consumed host and network resources. Typically, previous approaches utilize either IP multicast or point-to-point unicast for efficient dissemination of the information.In this work, we propose a novel hybrid framework, which is a combination of both multicast and unicast data dissemination. Our hybrid framework allows us to take the advantages of both multicast and unicast, while avoiding their drawbacks. We investigate several algorithms for computing the best mapping of publishers' transmissions into multicast and unicast transport.Using extensive simulations, we show that our hybrid framework reduces consumed host and network resources, outperforming traditional solutions. To insure the subscribers interests closely resemble those of real-world settings, our simulations are based on stock market data and on recorded IBM WebShpere subscriptions.

A hybrid architecture for delivery of panoramic video

“…The general idea is to publish tiles requested by many end-users on multicast channels and publish the less popular on unicast channels. The combination of multicast and unicast to optimally distribute content under capacity constraints has been studied in [17]. The application to the transport of tiled content has been studied in [18] in the case of using a shared medium (such as Wireless LAN).…”

On tile assignment for region-of-interest video streaming in a wireless LAN

“…The Hyper project uses a hybrid approach [14] where a multicast distribution tree may be bypassed in order to establish a unicast channel between the publisher and subscriber. Bickson et al investigated the mapping of flows into multicast and unicast sets in a quantitative manner [2] with heuristic solutions such as, random assignment with flow merges, use of k-means clustering, and use of binary matrix decomposition. In all these approaches a generic data type is assumed.…”

Modeling and optimizing content-based publish/subscribe systems

“…Another related approach of Bickson et al [11] intelligently combines IP multicast and point-to-point unicast for the efficient dissemination of the information. This leads to the so called channelization problem which is NP-Hard.…”