Project Overview

SuperNet

Interconnected set of DARPA-sponsored testbed networks

Purpose is to support networking and applications research

Access/transport ranging from 2.5Gb/s IP-over-SONET to end-to-end optical

Applications on SuperNet will:

• enable realistic testing of network technology
• demonstrate utility of high-performance networks
• provide end-users with state-of-the-art network capabilities
• motivate the design of network-capable and network-aware applications

SuperNet Testbed

DARPA is deploying a wide-area testbed, SuperNet, to support networking and applications research. SuperNet is comprised of several interconnected networks using technologies such as 2.5 Gb/s IP over SONET and end-to-end optical. The component networks:

• NTON II on the West Coast
• ATDNet/MONET in the Washington DC area
• BossNet between Boston and DC
• ONRAMP in the Boston area
• HSCC (high speed connectivity consortuim) providing access and cross-country connectivity

Return to top

Matisse Architecture

Testing of MEMS (micro-electro-mechanical systems) is an application that will run over the SuperNet.

MEMS co-locate, on single devices, sensing, actuating, and control functions along with computing and communications. MEMS build on micro-electronic manufacturing but are much more complicated because their fabrication involves many specialized steps, and testing must be performed on mechinical as well as electronic operation. Specialized platforms generate massive amounts of data that must be processed, stored, and visualized, and often ccmpared with tests on antecedent devices. Typically, a device designer from industry or academia will travel to a site that hosts one of the test platforms, may stay at the site for a week or more to ensure that the tests have been set up correctly and will then return home to wait for the test results, which may arrive in the form of a written report several weeks later.

By taking advantage of the capabilities of a high -performance network, the steps outlined above can be greatly compressed. The architecture that will be used is shown below. Data, originiating at the source such as a MEMS test platform, is sent first to a network-based, high-speed storage cache, the central component of the architecture. From the cache, the data can be sent to an archive for long-term storage, and/or to a computing facilty for processing. The processed data are sent to the user (directly, or via the cache) and/or to the archive. Previously obtained raw or processed data can be retrieved from the archive and re-examined, re-processed, or compared with new results. An important point to note is that the five architectural elements can be located anywhere because they are linked by a network. Moreover, if computing resources, for example, are available at several sites, then the system should be able to select the one that will give best performance (or lowest price). Further, there is no limit to the size of the dataset that can be handled, so long as the cache can temporarily store the volume of data currently needed.

The Matisse Project will implement this architecture for the MEMS R&D community.

Return to top

Matisse Platform

The components of the MEMS application running on SuperNet:

• a test platform at MIT in Cambridge, MA
• and the University of California, Berkeley
• the cache and the archive at Lawrence Berkley National Laboratory (LBNL)
• a multi-processor computing resource (developed by Sarnoff Corp.) at ISI-East in Arlington, VA
• and a user at CMU in Pittsburgh

Return to top