Project UNICORE Plus
Synopsis
Advances in High Performance Computing have provided science with a new research tool. Computational research and engineering used in combination with established methods of analysis and observation makes investigation of natural phenomena and technical problems possible beyond the reach of the traditional methods alone. Indeed, for macro-scale phenomena such as astrophysics and planetary weather, which cannot be controlled or reproduced in the laboratory, computational research may offer the only possibility for controlled experimentation.
Many researchers in industries and academic do not yet have easy access to the existing high performance facilities. It is economically not feasible to provide required systems to each research and development environment. On the other hand, interfaces to supercomputing resources over the network tend to be both complicated and vendor specific.
Goals
The goal of UNICORE Plus is to develop a grid infrastructure together with a computing portal for engineers and scientists to access supercomputer centers from anywhere on the Internet. This has to be done with strong authentication in a uniform and easy to use way. The differences between platforms will be hidden from the user thus creating a seamless interface for accessing supercomputers, compiling and running applications, and transferring input/output data.
Research areas in UNICORE Plus are resource modeling, application specific interfaces, data management, job control flow, and metacomputing.
Status and Results
The project is based on the work done in the UNICORE project.
UNICORE Plus started in January 2000 with the implementation of the revised architecture:
The system is fully implemented in Java with the exception of the target system interface (TSI) which is implemented in Perl.
During the project phase the certification authority (CA) which issues certificates for UNICORE had been established at LRZ. The UNICORE CA policy was based on the DFN-PCA WWW-policy .
UNICORE has been implemented for the target systems CRAY T3E and SV1 with NQS, Fujitsu VPP with NQS, Hitachi SR8000 with NQS, IBM SP and Regatta with LL, Origin 2000 with NQS, Origin 3000 with LSF, Siemens hpcLine with CCS and PBS, and others.
UNICORE is freely available at http://unicore.sourceforge.net .
Publications
last change 16.06.2010 | Sabine Höfler-Thierfeldt | Print
| Topic: | Uniform Interface to Computing Resources | |
| Granted by: |  BMBF
|
|
| Grant Period: | 01.01.2000 - 31.12.2002 | |
| Contact Person: | Dietmar Erwin, ZAM | |
| Partners: |
|
Central Institute for Applied Mathematics, Forschungszentrum Jülich GmbH |
|
|
DWD - Deutscher Wetterdienst, Offenbach | |
|
|
RUS - Rechenzentrum der Universität Stuttgart | |
|
|
Intel GmbH (formerly Pallas GmbH), Brühl | |
| Collaborators: |
|
LRZ - Leibniz-Rechenzentrum, München Germany |
|
|
PC2 - Paderborn Center for Parallel Computing | |
|
|
Rechenzentrum der Universität Karlsruhe | |
|
|
ZHR - Zentrum für Hochleistungsrechnen der Technischen Universität Dresden | |
|
|
ZIB - Konrad-Zuse-Zentrum für Informationstechnik Berlin | |
Synopsis
Advances in High Performance Computing have provided science with a new research tool. Computational research and engineering used in combination with established methods of analysis and observation makes investigation of natural phenomena and technical problems possible beyond the reach of the traditional methods alone. Indeed, for macro-scale phenomena such as astrophysics and planetary weather, which cannot be controlled or reproduced in the laboratory, computational research may offer the only possibility for controlled experimentation.
Many researchers in industries and academic do not yet have easy access to the existing high performance facilities. It is economically not feasible to provide required systems to each research and development environment. On the other hand, interfaces to supercomputing resources over the network tend to be both complicated and vendor specific.
Goals
The goal of UNICORE Plus is to develop a grid infrastructure together with a computing portal for engineers and scientists to access supercomputer centers from anywhere on the Internet. This has to be done with strong authentication in a uniform and easy to use way. The differences between platforms will be hidden from the user thus creating a seamless interface for accessing supercomputers, compiling and running applications, and transferring input/output data.
Research areas in UNICORE Plus are resource modeling, application specific interfaces, data management, job control flow, and metacomputing.
Status and Results
The project is based on the work done in the UNICORE project.
UNICORE Plus started in January 2000 with the implementation of the revised architecture:
The system is fully implemented in Java with the exception of the target system interface (TSI) which is implemented in Perl.
During the project phase the certification authority (CA) which issues certificates for UNICORE had been established at LRZ. The UNICORE CA policy was based on the DFN-PCA WWW-policy .
UNICORE has been implemented for the target systems CRAY T3E and SV1 with NQS, Fujitsu VPP with NQS, Hitachi SR8000 with NQS, IBM SP and Regatta with LL, Origin 2000 with NQS, Origin 3000 with LSF, Siemens hpcLine with CCS and PBS, and others.
UNICORE is freely available at http://unicore.sourceforge.net .
Publications
|
|
"UNICORE and the Project UNICORE Plus" Erwin, Dietmar (2000) Presentation at ZKI working group Supercomputing meeting on May 25, 2000 |
|
|
"UNICORE: Beyond Web-based Job-Submission" Romberg, Mathilde (2000) Proceedings of the 42nd Cray User Group Conference, May 22-26, 2000, Noordwijk |
last change 16.06.2010 | Sabine Höfler-Thierfeldt | Print