Teams of scientists from research organizations around the world competed recently in Phoenix to see who could move the most scientific data across networks in the fourth annual High-Performance Bandwidth Challenge, held in conjunction with SC2003, the international conference on high-performance computing and networking. SC2003 is sponsored by the Institute of Electrical and Electronics Engineers Computer Society and by the Association for Computing Machinery’s Special Interest Group on Computer Architecture.
This year’s winners included the following:
Sustained Bandwidth Award: “Bandwidth Lust: Distributed Particle Physics Analysis Using Ultra-High Speed TCP on The Grid.” A team from the Stanford Linear Accelerator Center, Caltech, Los Alamos National Laboratory and CERN moved a total of 6551.134 gigabits of data, reaching 23.23 Gbps.
Tools Award: “High Performance Grid-Enabled Data Movement with GridFTP,” which emphasized creating common, standards-based tools that are the building blocks for new applications, and demonstrating it capability with visualization. Sustained high rate was 8.94 Gbps by a team from Argonne National Laboratory and San Diego Supercomputer Center (SDSC).
Application Foundation Award: “DataSpace,” which used a Web service framework integrated with high-performance networking tools to provide an application foundation for the use of distributed datasets. High sustained rate was 3.66 Gbps by a team from the University of Illinois at Chicago, Northwestern University, University of Amsterdam; SURFNet, John Hopkins University and Oak Ridge National Laboratory.
Application Award: “Multi-Continental Telescience,” which emphasized user interaction with science instruments, distributed collaboration, with particular attention to ease of use by domain scientists. The team, which posted a sustained rate of 1.13 Gbps, included researchers from the University of California at San Diego, Universidad de Buenos Aires, Karolinska Institute, Osaka University, Center for Ultra High Voltage Microscopy, NCHC and SDSC.
Distance x Bandwidth Product & Network Technology Award: “Transmission Rate Controlled TCP on Data Reservoir, University of Tokyo,” which demonstrated attention to the details of controlling multiple gigabit streams fairly over extremely long distances. The team achieved a very high average pipe utilization of over 65% with real disk-to-disk transfer with a high sustained rate of 7.56 Gbps. Team members are from the University of Tokyo, Fujitsu Laboratories and Fujitsu Computer Technologies.
Commercial Tools Award: “On-Demand File Access over a Wide Area with GPFS,” showing emergence and use of commercial system that demonstrates high-performance without significant impact on remote systems. The team from SDSC and IBM posted a sustained rate of 8.96 Gbps.
Distributed Infrastructure Award: “Trans-Pacific Grid Datafarm,” a geographically distributed file system which took advantage of multiple physical paths to achieve high performance over long distances. The team achieved a high rate of 3.57 Gbps. Team members are from Japan’s National Institute of Advanced Industrial Science and Technology, Tokyo Institute of Technology, University of Tsukuba, KEK, APAN Tokyo XP and Indiana University.
Both Directions Award: “Distributed Lustre File System Demonstration,” which proved that not all applications or bandwidth challenge entries move data only in one direction. The team achieved a rate of 9.02 Gbps. Team members are from Cluster File System, Acme Microsystems: Supermicron; Foundry Networks, Data DirectNetworks, S2io; Sandia National Laboratories, Los Alamos National Laboratory, National Center for Supercomputing Applications, SDSC, and Lawrence Livermore National Laboratory.
A graphical representation of each team’s effort, along with detailed statistics on the amount of data transferred, can be found at http://scinet.supercomp.org/2003/bwc/results/index.html .
For the fourth consecutive year, Qwest Communications sponsored prizes for the winning teams.
http://www.sc-conference.org/sc2003/nr_finalbwc.html
