10.00 - 11.00 -Onno Boxma (NWO-GBE, Eindhoven Univ. of Technology) Opening -Ian Foster (Argonne National Laboratory) Open Grid Services Architecture 11.30 - 11.40 Henri Bal (VU) - Introduction DAS-2 11.40 - 12.00 John Romein (VU) - Retrograde analysis 12.00 - 12.30 Dick Epema (Delft) - Processor co-allocation 13.45 - 14.30 Daron Green (IBM) - IBM and Grid computing 14.30 - 15.00 Adam Belloum (Amsterdam) - Virtual laboratories 15.30 - 16.00 Herbert Bos & Lex Wolters (Leiden) - Education 16.00 - 16.30 Aad van der Steen (Utrecht) - Distributed data assimilation Program
Distributed ASCI Supercomputer 2 (DAS-2) Henri Bal Vrije Universiteit Amsterdam Faculty of Sciences
Distributed ASCI Supercomputer (1) Shared testbed for research on parallel and distributed computing -Set up by ASCI research school in 1997 -Geographically distributed cluster-based system -4 clusters with 200 nodes in total -Homogeneous (same processors, network, OS) Opening workshop: 2 March 1998 -Carl Kesselman, Andrew Grimshaw
Second testbed: DAS-2 ASCI obtained funding from N.W.O. in June 2000 ASCI committee to do detailed design and selection -Henri Bal, Bob Hertzberger, Henk Sips, Lex Wolters, Aad van der Steen, Cees de Laat Decisions -Homogeneous architecture -Nodes: dual Pentium-IIIs (SMPs) -Network: Myrinet + Ethernet (OS) -OS (Linux) -Vendor (IBM), configuration, …. Installation by IBM: November 2001
DAS-2 VU (72)Amsterdam (32) Leiden (32) Delft (32) GigaPort Utrecht (32) 200 SMP nodes (IBM xSeries) with two 1GHz Pentium-IIIs, 1 Gbyte memory
Systems research in ASCI Distributed supercomputing (Albatross) Mapping data-parallel programs (AUTOMAP) Compilers for distributed/embedded systems (JOSES) High-performance Java (Manta) Middleware for scalable wide-area applications (Globe) Distributed web caching Dynamic task migration (Dynamite) Resource management in wide-area systems Quality of Service Authentication, Authorization, and Accounting (AAA)
Applications research in ASCI Virtual laboratories (VL) Image processing (PILE) Weather forecasting (Hirlam) Search in large image databases Ubiquitous communications (UBICOM) N-body simulations (GRAPE) Web servers Autonomous interacting robots (AIR)
DAS-2 Application performance Case study: endgame databases for Awari 1995 (Bal & Allis, Supercomputing’95): -Computed database with 5 10 8 entries on 80-node MicroSparc cluster (predecessor of DAS-1) 2002 (Romein & Bal): -See how far we get on DAS-2 towards solving the entire game (9 10 12 entries) ….