The Tau Project

Tau is motivated by what has been learned about end-to-end protocol design and implementation in the past ten years. Although present practice reflects some of this (congestion control, for example), it is well known that some of these techniques --for example, Integrated Layer Processing (ILP)-- are hampered by the use of layering as a composition mechanism above TCP and UDP. Much of the standards activity today is at the upper levels (cf. SSL, HTTP, DNS, etc), yet the Internet basically has no architecture above TCP and UDP ports. Layering functionality on top of TCP and UDP in the traditional ways will increasingly limit performance, especially as round trips inevitably become the dominant performance consideration. The primary objective of the Tau project is to create a standard platform for development and deployment of new protocol services, one designed to incorporate what has been learned about protocol architecture in the past 20 years.

The Tau framework comprises three major parts:

• An explicit metaheader protocol, which provides application addressing and an indication of the protocol processing applied to the message. The metaheader is coded to allow efficient parsing and processing of Tau data units by the receiver.
• A generic protocol interface (GPI), which allows protocol functionality to be specified and implemented apart from the ``glue'' that binds them together.
• An extensible set of protocol functions, including connection management, congestion avoidance and control, and security functions.

Tau can be instantiated with different "top" and "bottom" interfaces, so it can be used with different underlying services. It is designed to be used over UDP, TCP, or directly on IP; both user-space and kernel implementations are possible. The basic design premise of Tau is that it should be possible to maintain modularity of implementation (and thus increase code reusability) and still support high-performance implementation techniques, by using an explicit composition mechanism instead of layered encapsulation.

Tau specification

Publications Related to Tau

• R. Clayton and K. Calvert, "A Reactive Implementation of the Tau Composition Mechanism", Proceedings of the 1998 IEEE Conference on Open Architectures and Programming (OpenArch 98).

• Joseph T. Dixon and Kenneth L. Calvert, "Effective Search Strategies for Application-Independent Speedup in UDP Demultiplexing", Proceedings of the Internation Conference on Computer Communication Networks, Las Vegas, September 1997. Longer version available as College of Computing Technical Report GIT-CC-97-02 .

• Joseph T. Dixon and Kenneth L. Calvert, "Increasing Demultiplexing Efficiency in TCP/IP Network Servers", Proceedings of the Internation Conference on Computer Communication Networks, Washington, D.C., October 1996. Longer version available as College of Computing Technical Report GIT-CC-96-08 .

• Richard V. Clayton and Kenneth L. Calvert, "Structuring Protocols as Data Streams", Second Workshop on High-Performance Protocol Architectures, Sydney, Australia, December 1995.

• Kenneth L. Calvert, "Beyond Layering: Modularity Considerations for Protocol Architectures", Proceedings 1993 International Conference on Network Protocols, San Francisco, October 1993. (gzipped PostScript)

• Robin Kravets, Ken Calvert, and Karsten Schwan, "Dynamically Configurable Communication Protocols and Distributed Applications: Motivation and Experience", College of Computing Technical Report GIT-CC-96-16
  .

Students

• Tom Browning
• Richard Clayton
• Joe Dixon
• Robin Kravets
• Bobby Krupczak

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.