A Hypercubic Overlay Using Bloom-Filter Based Addressing for a Non-dedicated Distributed Tag-Based Pub/Sub System

The proliferation of Web 2.0 technology, coupled with the advancements in social networking, have unraveled a new Internet information exchanging paradigm where every individual user involved is capable of publishing information at will and at any time, while also being able to filter and selectively receive any information they are interested in a timely manner. To facilitate this process, users are allowed to tag their published content with one or more keywords (namely tags), and submit their interests represented by a conjunction of tags. The content will be dispatched to users whose interests matching the content's tags. Publish-Subscribe, in short Pub/sub, communication paradigm suits perfectly for this scenario, by providing loosely coupled and content-oriented communication among parties. To date, existing implementations of tag-based pub/sub are mostly based on polling, which requires the client to constantly query the server for the update in a pull-based manner. Thus the essence of pub/sub, which exercises decoupling between communication parties, could not be fully exploited. Established researches on topic-based and content-based pub/sub system point to the right direction of adopting fully distributed push-based architecture. Especially, the Key-based Routing (KBR) overlay-based approach, which supports a large-scale dissemination of messages with dynamic topology changes, is highly preferable for a non-distributed environment. However existing works on KBR based approach relies on static categories/topics or fixed filter dimensions in the underlying matching schemes, which make them unsuitable for tag-based systems that require support of variable filter matching. This motivates us to find a way to utilize KBR in a tag-based pub/sub system. In this research, we adopt a holistic approach of treating matching and routing as a whole and proposed a scalable and robust hypercubic Publication-Dissemination overlay infrastructure for tag-based pub/sub system. With this approach, we provide a robust yet efficient multilayer non-dedicated distributed pub/sub architecture, which comply with the heterogeneity and scalability requirement of large-scale Internet applications.