Home  |  NARL  | 中文 
Sun, September 24, 2017

| High-performance Computing

Recent improvements in microprocessors and networking have made it possible for PC clusters to compete with conventional mainframes in supercomputing. In fact, cluster architecture, the current trend in supercomputing, continues to dominate with 84% of the supercomputers on the TOP500 supercomputing list being cluster computers. Based on its expertise in the field, the NCHC's PC Cluster research group built the Formosa series clusters and wrote the administrative software that operates them. This hardware/software combination creates a HPC platform that is easy to access and develop user applications over. The NCHC also provides its expertise in cluster computing and parallel processing for research groups via training courses and direct consultation.

Core technologies

  • PC Cluster auto installation and deployment
  • Cluster Computing on-demand and Dynamic Power Management
  • Parallel and heterogeneous programming
  • Parallel program tuning and optimization

Services offered

  • HPC Cluster middleware and system environment installation services
  • HPC Cluster professional training courses
  • HPC Cluster middleware performance tuning services
  • HPC Cluster solutions consultation
  • Application tuning and optimization

Main achievements

  • Beginning in 2003, the NCHC started designing and building the Formosa series of cluster computers. The Formosa cluster computers possess extraordinary parallel cluster technology and serve to demonstrate the NCHC's technological know-how that it has accumulated over many years.
  • Parallel CDR Library: The Parallel CDR Library is a set of code for solving a convection–diffusion–reaction (CDR) scalar transport equation that uses GPU cards to obtain results in the shortest amount of time. The main purpose of the pCDR Library is to make GPU programming easy for CFD scientists.
  • SimKernel: the simKernel module provides a user-friendly environment for using large-scale solvers on NCHC's supercomputer ALPS.
  • Large-scale solvers with auto tuning parameter selection.

Contact information
Cluster Computing: Shuen-Tai Wang, 886-6-5050940 ext.649 * stwang@narlabs.org.tw
Parallel Computing: Chau-Yi Chou, 886-4-24620202 ext.828 * cychou@narlabs.org.tw

Home | Top | Back