 |
Services Offered |
|
|
|
|
|
|
|
Core Technology |
|
|
|
|
|
About Us |
|
|
|
|
|
|
|
 |
PC Cluster Computing
|
Introduction
|
Although the history of the PC Cluster is short, its growth has been phenomenal! Due to its combination of high computational capability and affordability, more and more researchers have converted to the PC Cluster. Because of academia and industry's increasing demand for PC Clusters, in 1999, we formed a PC Cluster workgroup to actively research, develop, and promote this technology. Due to the PC Cluster's relative low-cost and high computational capacity, it is the practical choice for Taiwan's existing and foreseeable supercomputing needs.
Recently, we constructed two PC Clusters to support domestic R&D in the science and engineering fields. Our PC Clusters provide researches an outstanding testbed environment on which to develop new software applications. Our PC Cluster’s performance totals out at 938.2 Gflops (R peak). In fact, the computing capability of our PC Clusters surpasses that of our individual IBM P690 and the HP Superdome supercomputers! Our extensive experience in the development of PC Clusters allows us to take full advantage of our cluster hardware resources. Additionally, we have helped many other domestic institutes build their own PC Clusters at practical costs.
In 2004, we made several significant upgrades to our HP LC6000 PC Cluster, the cluster that manages our Knowledge Information National Grid (KING) Knowledge Management Center. Our upgraded HP LC6000 cluster now consists of 192 HP integrity rx2600 servers and 384 Intel Itanium 2 1.5 GHz processors. It also utilizes Quadrics ELAN4 for high-performance interconnectivity and is capable of reaching computational speeds of up to 2.0 TFLOP/s. This was Taiwan's first 64-bit cluster. Utilizing this cluster, we will be able to meet the demands of our research projects requiring leading-edge distributive calculation capabilities and we will be able to more effectively participate in Grid computing endeavors with other supercomputing centers domestically and abroad.
|
The Goals of Our PC Cluster Program Include:
|
- Bringing in, developing, and promoting cluster and Grid computing-related technologies and applications,
- Building a computational Grid testbed by connecting, via the Internet, PC Clusters located at various Taiwanese universities,
- Strengthening and evaluating parallel processing techniques for various application domains,
- Developing management and queuing software for the cluster & Grid environments and expanding our software development team,
- Enhancing our consultation services, and
- Setting up an on-line cluster computing forum (http//pccluster.nchc.org.tw) and offering HPC-related courses (i.e. Cluster computing training workshops).
|
Our 2005 PC Cluster-Related Experience and Performance:
|
- Constructed PC Clusters to support 200+ users and 100+ projects.
- Optimized the Formosa PC Cluster; ranked #135 on the 11/2003 TOP500 supercomputer list.
- Published 20+ professional papers in scientific journals on PC Cluster and parallel computing.
- Developed the Cluster Management Enabling Toolbox for Administrators (Cmeta) software tool; 500+ downloads at http://pccluster.nchc.org.tw/download.
- Conducted the Cluster computing training camps; 2,000+ attendees.
- Maintained the on-line Cluster computing forum; 1,100+ registered members, 2,300+ discussion articles, and 32,000+ hits.
|
Parallel Computing
|
The vector supercomputers of the 80s and 90s are slowly being phased out. They are being replaced by massively parallel computing systems and PC Clusters. Effective parallel programming has become critical to the progression and development of parallel computation. In order to help advance the field of parallel programming, the NCHC provides parallel programming and computation workshops as well as consultation. In 2005, we held a total of three two-day-long MPI parallel programming courses. We also consulted with Professor Niu of Chung Hua University and Professor Jang of National Kaohsiung Institute of Marine Technology to help develop their parallel application programs.
|
|
TOP | BACK |
|
|
| Challenge 2008 |
| |
|
| |
|
| |
|
| |
|
|
| |
| Professional Knowledge |
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| Community Outreach |
| |
|
| |
|
| |
|
