[Acpc-l] AURORA: Vortrag: Fr., 04.10.02 - 14:00 h s.t. - Dr. Randy Moe (Univ. of Bergen)
AURORA Project
AURORA Project <aurora@par.univie.ac.at>
Wed, 2 Oct 2002 13:50:11 +0200 (MEST)
EINLADUNG ZU EINEM VORTRAG IM RAHMEN DES AURORA-KOLLOQUIUMS
UNIVERSITAET WIEN INSTITUT FUER SOFTWAREWISSENSCHAFT
gemeinsam mit
FWF-Projekt Spezialforschungsbereich F011 "AURORA"
EINLADUNG ZU EINEM VORTRAG IM RAHMEN DES AURORA-KOLLOQUIUMS
ZEIT: Freitag, 4.Oktober 2002, 14 Uhr s.t.
ORT: Institut fuer Softwarewissenschaft,
1090 Wien, Liechtensteinstrasse 22,
Seminarraum, Mezzanin
Dr. Randy Moe
University of Bergen
Department of Informatics
Norway
GRIBB - an infrastructure for Internet computing
Abstract:
With the introduction of the URL and HTML standards for defining
unique document addresses and links between them, the ground was
prepared for wide-spread use of the Internet.
When the appropriate software, the web-browser, came along,
the use of Internet as an information channel exploded.
Internet based computing seeks to exploit otherwise idle workstations and
PCs to create powerful distributed computing systems with global reach and
supercomputing facilities. Different systems have been introduced to harness
idle cycles on computers for useful work both on local networks as well as
globally through the Internet.
The GRIBB project is based on an ongoing project called GRISK which was
started for solving a particular problem, that of finding K-optimal lattice
rules for numerical integration using an Internet based search. Hundreds
of computers from more than 20 different countries have been used
and more than 100 CPU-years have been consumed to produce new scientific
results which would have been beyond reach without Internet based computing.
The software written to accomplish this is very compact and efficient
making use of tools like JAVA and RMI (Remote Method Invocation).
In the GRIBB project we build on these experiences, applying the same basic
philosophy and core technology to a wider range of problems - namely
branch-and-bound methods.
Compared to similar projects GRIBB has several advantages like robustness, a
very simple structure, and very little overhead. The GRIBB software is not
dependent on any other software components than JAVA and the problem solving
code.
The Quadratic Assignment Problem has been chosen as the first major test
application due to its high computation complexity, and preliminary results
from test runs will be presented.