![]() |
This 128-node dual-XeonHPC system, which has been given the host-name of "hpcpower.hku.hk", has a theoretical peak processing speed (Rpeak value) of 1433.6 GFlops, the fastest in Hong Kong. It is about 20 times more powerful than the existing hkuhpcc system in the Computer Centre, and has just been ranked the 240th position on the world-wide Top 500 Supercomputer list released in November 2003.
Software installed on this system include BLAS mathematical library, Gaussian 03 with LINDA, PGI compiler suites for C, C++, Fortran 77, Fortran 90 and High Performance Fortran, PBS job queue system and the Maui job scheduler. System software in the hpcpower system is similar to that in the hkuhpcc system, but the user file directories are completely independent of the hkuhpcc system.
For program development tasks, including program editing, compilation, testing and debugging, that need to be done interactively through telnet sessions, users can login to the hpcpower.hku.hk system directly. Since it is expected that running jobs interactively by the users should be mainly for debugging purpose, all telnet sessions are served on the master node only and the maximum CPU time for each interactive background jobs will be limited to 15 minutes.
Batch Job Submission
For longer runs of program jobs that require more than two CPUs and longer processing time, they must be submitted as batch jobs through the PBS job queuing system.
Since the system is set up to support large computation jobs, the following maximum number of CPUs and maximum processing time are allowed for each batch job::
- Maximum number CPUs for each program job
= 64 (i.e. 32 compute-nodes of 2 CPUs)
- Maximum processing time for each program job =
10 Hours (wall clock time)
Furthermore, the job scheduling is set in such a fashion that higher priority will be given parallel jobs requiring a larger number of CPUs.
For any enquiry, please contact the undersigned.