DOE conducted a series of workshops during 2002-2004 to identify the networking needs of its science applications, and to chart a roadmap for the needed networking capabilties. It was felt that the needed high and stable bandwidths for large-scale science applications were not provided by the exisiting networks. In fact, such capabilties were deemed to be very hard to achieve over shared IP networks in general, and therefore there was a need for "other type" networks that can directly provide stable, unimpeded bandwidth to applications. Such networks were not available at that time except in very limited laboratory settings. Furthermore, the component technologies needed to develop these networks, such as user access to control planes, signaling techologies, and on-demand/in-advance provisiong technologies, were only partially available.
USN was conceived as a testbed to facilitate the development of the networking technologies that are capable of providing dedicated bandwidth connections to applications through on-demand or in-advance scheduling. Such networks enable the user to control the network connections, which is a radical departure from the shared IP paradigm. The technology development needed for such networks required access to switches and routers, which was not practical in production networks. USN provides such an access and furthermore, provides access to its hosts that can be used for the application development and testing.
USN was conceived and proposed to DOE in 2003, with first round of funding provided in FY2004. It has been fully deployed in 2005 in terms of the backbone. Its control plane is being finalized and peering technologies are being developed to interface it with CHEETAH, HOPI, OSCARS and DRAGON control planes. USN has also been used to test theconnectivity to ORNL Cray X1 supercomputer, and future plans include focussing on effective dedicated connectivity to supercomputers in general. As the networking components of USN are being matured, its scope is being expanded to include connectivity to suspercomputers, testing remote file and storage systems, and testing high-performance security components.