ShapeShifter is a content delivery system based on end-system multicast and erasure codes. It features improved overlay management, adaptivity, and reliability.

ShapeShifter uses erasure encoding to allow a rich connectivity graph, allowing the retrieval of content from multiple participating nodes. It employs sophisticated techniques to ensure that erasure-encoded content supplied retrieved from multiple nodes is not redundant.

ShapeShifter achieves scalability by its nodes needing only be aware of a small portion of the graph, while achieving graph coverage through continuous discovery.

ShapeShifter conducts and utilizes measurements of metrics most crucial to the optimization of the overlay graph, such as shared-link congestion.

The pressing problem of efficient delivery of popular bulk content is commonplace on the contemporary Internet. Bulk content, whether in the form of software patches, demos or multimedia files, comprises most of the Internet traffic by volume, and elaborate strategies are required in order to ensure reliable delivery and satisfactory performance.

The proposed approaches to bulk content delivery include schemes for optimized single-source unicast, forward caching/Content Delivery Networks, Reliable IP Multicast, and End-System Multicast.

Network-supported (IP) multicast is a service model which presents opportunities for near-optimal solutions. However, it lacks real-world deployment, suffers from scalability issues, and requires a separate solution to achieve reliability.

On the other hand, end-system multicast does not require network support, as it builds and manages a virtual overlay topology of unicast links on top of the network's physical topology. It also provides flexibility of optimization of the tree according to a richer set of metrics (perhaps specified by the application), and ability to change topology on-demand. On the downside, end-system multicast requires far more network resources than IP Multicast, due to the possibility of non-optimal mapping of virtual topology to physical network topology.

Principal Investigator:  John Byers

Core Team:

Jeffrey Considine
Michael Mitzenmacher
Stanislav Rost

Associates/Former Members:

Nicholas Eskelinen
Khaled Harfoush
Dmitriy Zavin


This project is funded by NSF CAREER Award ANI-00939296 (July 2000) and by NSF Research Infrastructure Award CNS-0202067 granted to Boston University. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
ShapeShifter is being developed at the Computer Science Department at Boston University.

For the purposes of testing, ShapeShifter will be deployed in both the local "Internet-in-the-lab" and the wide-area "lab-in-the-Internet" environments.

The project is developed from the ground up using GNU C++ on Linux. The development takes place both in user-space and kernel-space. In particular, we use a customized version of the Congestion Manager to achieve flow control. The codebase of ShapeShifter comprises a powerful framework which should drastically reduce the development time of future large-scale projects.
Narada/End-System Multicast   Overcast   Resilient Overlay Networks   RMX   Yallcast/Yoid
NSF visit slides: 2001 (PowerPoint) 2002 (PowerPoint)