1. e-VLBI – Creating a Global Radio-Telescope Array via High-Speed Networks Alan R. Whitney MIT Haystack Observatory Internet2 Fall Member Meeting San Diego, CA 11 Oct 2007

2. Traditional VLBI The Very-Long Baseline Interferometry (VLBI) Technique (with traditional data recording) The Global VLBI Array (up to ~20 stations can be used simultaneously)

3. VLBI astronomy example ASTRONOMY Highest resolution technique available to astronomers – tens of microarcseconds Allows detailed studies of the most distant objects – quasars, gravitational lenses, GRBs; as well as black hole at center of Milky Way PRECISION GEODESY Highest precision (few mm) technique available for global tectonic measurements Highest spatial and time resolution of Earth’s motion in space for the study of Earth’s interior Earth-rotation measurements important for military/civilian navigation Fundamental calibration for GPS constellation within Celestial Ref Frame VLBI Science Plate-tectonic motions from VLBI measurements

4. VLBI Data Rates and Volume Astronomy experiments at 1-4 Gbps/station, 4 to 20 stations –~5-40 TB/station/day –Global 10-station experiment @ 4 Gbps/station  up to ~400 TB/day –Single 10-day experiment can produce up to ~4 PB Higher data rates (8-32 Gbps) are already on the horizon Available disk supply can support only few days of observations at these rates All pairwise telescope combinations must be cross-correlated Uniqueness of VLBI data Transmitted data are uncompressible white Gaussian noise; useful signals appear only after correlation processing Loss of some data is tolerable “e-VLBI” is a natural fit to high-speed global networks!

5. Why do e-VLBI? Pre-experiment verification and diagnosis Rapid processing turnaround –Astronomy Quick feedback for adjustment of observations –Geodesy More timely Earth-orientation measurements; important for precision military and civilian navigation Bandwidth growth potential for higher sensitivity –VLBI sensitivity is proportional to SQRT(bandwidth), limited by availability of media –e-VLBI potential bandwidth to 100 Gbps/station or more Elimination and/or reduction of expensive media pool

6. The Colloquial Competition – a B747 loaded with 1 TB disks! Payload: 140 tons ≈ 140,000 disks = 140 PB Based on 24-hr flight time, bandwidth is ~10 Tb/sec! Cost estimated at $250,000 per flight! Note: In 1970, with 12” open-reel computer tape at 800 bpi, B747 bandwidth was only ~100 Mbps – down a factor of 100,000 from today!

7. Clearly, e-VLBI is possible and attractive, but there are a few problems…… Not all antennas are connected; many are in remote locations – high capital cost If even a single antenna cannot deliver data to correlator, all data must be recorded Shared networks are unable to deliver deterministic and repeatable performance to demanding applications (such as e- VLBI) –Unpredictable conditions within the network itself –TCP protocol congestion response can cause major data-rate slowdowns –UDP is not TCP friendly and is frowned upon in most shared networks

8. International e-VLBI Demonstrations iGRID-05 Networking Conf. San Diego, CA Sep 05 SC-05 Supercomputer Conf. Seattle, WA Nov 05 Internet2 Driving Exemplary Applications (IDEA) Award to “Very High Speed VLBI (e-VLBI)”. Alan Whitney (MIT/Haystack), Arpad Szomoru (JIVE), Y. Koyama (NICT), and Hisao Uose (NTT) Apr 26, 2006 – Arlington, VA

9. 155 Mbps 1 Gbps

10. National Astronomical Observatory of Japan (NAOJ) Kashima Space Research Center (CRL) NTT Musashino R&D Center National Institute of Information and Communication Technology (NiCT) KSP Koganei Station KSP Kashima Station e-VLBI Network in Japan Usuda Deep Space Center (ISAS) Nobeyama Radio Observatory (NAOJ) JGN2 Gifu University GSI SINET3 Yamaguchi University SINET3 GEMnet2/NTT PW/Internet2

11. Connectivity in Australia – current and planned

12. VLBA – Very Long Baseline Array Mauna Kea HI Owens Valley CA Brewster WA N. Liberty IA Hancock NH Kitt Peak AZ Pie Town NM Fort Davis TX Los Alamos NM St. Croix Virgin Is.

13. Collaborations e-VLBI is a highly collaborative global program Collaborations with: –NASA Goddard Space Flight Center –University of Maryland/ISI-E/GMU DRAGON project –Caltech High Energy Physicists under the leadership of Harvey Newman –Caltech Netlab Department –R&E Networks: Internet2, Dante, SURFnet, APAN, NORDUnet, SUNET, SINET –JIVE, The Netherlands –Jodrell Bank, England –Max Planck Institute, Germany –Shanghai Radio Observatory, China –CSIRO, Australia –NICT, Japan

14. e-VLBI Challenges Connect the telescopes to high-speed networks –Proceeding well in Europe, Japan, Asia and Australia –Many U.S. telescope are very isolated; very expensive to connect Find schedulable, reliable global high-speed connections –On-demand dynamic circuit networks are ideal, and probably required at higher data rates –HOPI and DRAGON hybrid networks have been first step –Challenge to set up across multiple international domains Need quick identification and resolution of failure points across multiple international domains Distributed cross-correlation processing must be developed to eliminate extreme data-rate concentration at correlator; some work is in progress Cost of doing e-VLBI must be competitive recording/shipping

15. So ……. The Bottom Line! e-VLBI has tremendous potential to improve both the science output and the rate of science output from global VLBI observations e-VLBI is probably the only practical way to extend data rates significantly beyond 10 Gbps/station Progress is being made, but significant investment must be made to fully realize potential The objectives of global CyberInfrastructure initiatives to provide on-demand, high-bandwidth, dedicated circuits across multiple global domains are key to long-term success. Thank you!