SemCast:
Semantic multiCast
Content-based Data Dissemination using
Multicast
Project
Overview
SemCast addresses the
problem of content-based dissemination of highly distributed data
streams for stream-based applications.
SemCast provides a selective distributed information dissemination
service, where routing is performed
based on the content of the messages (content-based
routing) rather than the IP addresses of the destinations. This
dissemination service follows the publish/subscribe paradigm,
decoupling
data sources from the subscribed clients. The goal of SemCast is to
efficiently identify and route relevant data to each client. The key
idea is to split the data streams based on their content,
the overlap of the clients's subscriptions, the stream rates and the
clients location, and spread the pieces across multiple semantic multicast channels. This
approach can lead to significant improvement in the consumption of
processing resources and bandwidth.
The key features of
our approach are the following :
- Efficient
content-based
filtering: SemCast requires filtering only at the sources
and the final destinations, eliminating the need for hop-by-hop
content-based filtering performed in traditional content-based
publish/subscribe systems.
- Flexibility:
SemCast disseminates each sub-stream through an independent
dissemination channel and thus, it facilates fine-control over the
construction of the multicast trees.
- Adaptivity:
SemCast continuously monitors the system's characteristics
and adapts to changes in the overlap among the subscriptions, the rate
of the incoming streams and the location of the clients by reevaluating
its channelization decisions.
The high-level design and
basic performance of SemCast is presented in an ICDE'05
paper. This paper subsumes
our WebDB'04 paper that provides a preliminary
desription of the system. This project is superseded by the XPORT project.
Publications
People
Current Members
Part Members