Thoughts on Objectives for Testbed Interconnections

T. J. Salo

Draft: 5/20/98

When we implement interconnections between MAGIC-II and other wide-area testbeds, some interconnection techniques seem to us to be "better" than others do. However, the superiority of a particular solution often appears to be highly dependent upon why one believes testbeds ought to be interconnected. Unfortunately, as far as I know, there is no consensus about the motivation for interconnecting testbeds, hence no agreement about why some techniques are better than others are. This note is intended to stimulate discussion within the testbed and network research communities about what the objectives for testbed interconnections ought to be. These objectives can then be used to support criteria for evaluating specific interconnection techniques and proposals.

I believe that the primary objective for a testbed interconnection ought to be to:

• Stimulate the development and demonstration of scalable interconnection solutions that provide the functionality required to construct the Next Generation Internet (NGI), Internet2 and other next-generation networks.

I believe that the following objectives are subordinate to, but collectively supportive of, this primary objective:

• Enhance the quality of testbed-based network research results by:
• Stimulating interaction and collaboration among testbed participants;
• Providing larger, more heterogeneous environments in which to develop empirical network research results;
• Promoting earlier deployment of research prototypes in larger, more heavily used operational environments; and
• Increasing the focus of newly developed network solutions on interoperability and scalability.
• Increase the return on the substantial investments made by government and industry in the testbed programs by:
• Creating larger, more diverse environments for application and network research than could otherwise be afforded;
• Increasing the utilization of the testbed assets by making larger, inter-testbed experiments easier to conduct; and
• Fostering an environment where it is easier for new research programs to use existing individual or interconnected testbeds.

I propose these objectives without extensive justification, in large part because I believe these objectives ought to result from collective discussions. I am hopeful that this note will help stimulate those discussions.

http://www.msci.magic.net/
Updated: 5/20/98