Vortrag: Dr. Ralf Gruber, 28.6.1999

Maria Cherry Maria Cherry <maria@par.univie.ac.at>
Mon, 21 Jun 1999 12:09:09 +0200 (MET DST)


                             UNIVERSITAET WIEN 
              INSTITUT FUER SOFTWARETECHNIK UND PARALLELE SYSTEME
                                gemeinsam mit 
                                    VCPC 
               EUROPEAN CENTRE FOR PARALLEL COMPUTING AT VIENNA 
                       
              FWF-Projekt Spezialforschungsbereich F011 "AURORA"


        EINLADUNG ZU EINEM VORTRAG IM RAHMEN DES AURORA-KOLLOQUIUMS
                         
            
                    
                       From Commodity to Supercomputers
  
			  
		              Dr. Ralf Gruber		            
                                 SIC-EPFL
                      CH-1015 Lausanne, Switzerland
		 
                  
                  ZEIT: Montag, 28. 6. 1999, 17.15 Uhr s.t.
          ORT: Institut fuer Softwaretechnik und Parallele Systeme
                   1090 Wien, Liechtensteinstrasse 22, 
                          Seminarraum, Mezzanin



Abstract:
EPFL has decided to replace the Cray-T3D/256 by a parallel machine
mainly consisting of commodity parts and called Swiss-T1. This machine
is built in the common Swiss-Tx project between EPFL, ETHZ, CSCS (Swiss
Center for Scientific Computing in Manno) and the industrial partners
Compaq and Supercomputing Systems (SCS) in Zurich. In this R&D project,
SCS builds the high performance communication system T-NET with a 12x12
crossbar, a PCI adapter and, together with ETHZ, a MPI communication
library based on the hardware implemented Fast Communication Interface
(FCI). EPFL designs the machines and contributes with its experience in
the management and usage of massively parallel computers. CSCS takes
care of parallel tools. 

The new generalised K-Ring architecture design guarantees that the
minimum maximal distance between computational nodes is small and the
routing optimal for global communications. At the time of the
presentation, it is expected that at least one computational node with
12 processors will be operational at EPFL and first experiences made.

The efforts that are made at EPFL to understand the relation between
parallel algorithms and the corresponding communication needs will be
discussed and the distributed data management system MEMCOM developed at
EPFL and especially at the commercialising company SMR SA presented.

For the Swiss-Tx project, see http://capawww.epfl.ch/
For MEMCOM, see http://www.smr.ch/