| Summary: | The Publish/Subscribe model has become a prevalent paradigm for building
distributed notification services by decoupling the publishers and the subscribers from each other. Content-based publish/subscribe allows for highly
expressive descriptions of subscriptions and thus is more appropriate for content
dissemination when a finer level of granularity is necessary. However,
scalability has become an issue due to the expensive matching and delivering
inherent in content-based events. In this thesis we propose a novel content based
publish/subscribe framework built over a DHT-based P2P network in
order to provide scalable content delivery mechanisms. Based on efficient
subscription installation, event publishing and event delivery techniques, our
system can provide a scalable platform to support multiple different pub/
sub schemas. There are three key features in our design: (1) A logic space
mapping and a distributed 2d-tree that maintains this space over DHT;
(2) Novel random probing searching schemes allowing for subscription installation
and event publication; (3) An efficient application layer multicast
algorithm for message delivery with low bandwidth consumption. === Science, Faculty of === Computer Science, Department of === Graduate
|
|---|
