Approach: The two ACTS experiments, Keck and GCM, are led by JPL and GSFC, respectively, with support from Caltech, UCLA, GWU, and Hawaii PacSpace. The GCM experiment requires a virtual channel connection between the JPL Cray T3D and the GSFC Cray C90, while the Keck experiment requires a virtual channel connection between a remote control room at the Caltech campus in Pasadena, CA, and the Keck Observatory local area network on Mauna Kea, Hawaii. The network transport protocol is based on Asynchronous Transfer Mode (ATM). Terrestrial network physical fabric (outside of ACTS) is based on fiber optic cables employing Synchronous Optical NETwork (SONET) OC-3 transmission as well as DS-3 microwave. ATM was selected as the base transport mechanism because of the diverse physical fabrics and bit error rates. This greatly simplifies the terrestrial network infrastructure, especially in the Hawaiian islands and ATDnet.
For Keck, Caltech modified the graphical user interface (GUI) design for use over longer delay channels and multiuser/location control (an adaptation of one currently used), JPL performed the network system engineering and atmospheric/fading BER analysis, and GWU the HDR site design and performance modeling. Additionally, PacSpace assisted with scheduling the use of the Honolulu HDR and engineering the Honolulu/Mauna Kea network infrastructure.
For GCM, GSFC led the porting of the distributed global climate model to the JPL and GSFC Cray supercomputers. GSFC staff scientists ported the Poseidon OGCM and Aries AGCM codes for coupling with UCLA AGCM and GFDL OGCM codes.
In both experiments, the effect of fading, burst noise, and long transit delays were examined and compared against lower error rate terrestrial links.
Accomplishments: During the past year (Oct 95 - Sep 96), both Keck and GCM experiments were successfully run over ACTS. One common upgrade to both experiments was modifying TCP/IP to extend the window size from 64KB to 2.5MB (this was done on the Crays for GCM and Sun SPARC 20's for the Keck) so that maximum channel capacity could be achieved. Also, since Keck employed interactive X-windows interfaces, some additional tuning of the window rampup time constant was also implemented to decrease response times. Also, a new Ximage module was written as a work around to the higher BER experienced on the microwave link segment in Hawaii which caused frame dropouts. Keck achieved 15Mbit/s throughput between Sun SPARC 20's over a DS-3 (45 Mbit/s) ACTS circuit with a TCP window size of 1MB (note Sun window memory was limiting factor), and GCM achieved 23 Mbit/s between the JPL Cray and GSFC Sun SPARC 20 over an OC-3 (155 Mbit/s) circuit using a 4MB window size.
Significance: This pair of experiments demonstrated the feasibility of using long path delay satellite links to establish meta-computing and control/data acquisition networks for remote collaboration, observation, and control of science experiments in hostile environments. Examples include Antarctic and undersea exploration, petroleum exploration, and interconnecting data centers to share large data bases.
Status/Plans: Although the Keck experiment milestone has been successfully completed in FY96, user trials are expected to continue through the end of FY97 when the interisland fiber trunk lease expires. During this period, additional telescope X-window modules are expected to be optimized, and several astronomers may use the link for impromptu investigations. Ultimately, the tools developed for this application will pave the way for interactive remote observing when high speed undersea trunk cables eventually become affordable. NASA is also investigating networking major observatories together in the future in this fashion.
Point of Contact:
Larry A. Bergman
Jet Propulsion Laboratory
(818)354-4689