You are here

9. Computer and Research Support Facilities Primary tabs

Powerful central computer servers support various web-based applications, departmental web sites and electronic email.

ITS provides the expertise and central support in high performance computing (HPC), high throughput computing(HTC) and grid computing to facilitate computational and data-intensive research activities of the University. Software packages for supporting engineering computation, scientific applications, simulations, graphics applications and statistical analysis are installed on these systems for use by students and staff.

The high performance computing (HPC) clusters utilize multiple processors for solving cutting-edge research computational problems and complex scientific simulations. A new 2,320-core HPC system (theoretical peak power of 103.4 Tera-FLOPS) equipped with the latest technologies, including GPGPU, MIC and Luster file-system, has just been put into service for supporting compute-intensive and accelerated computing.  In addition, a 2,256-core 64-bit HPC system (theoretical peak power of 29.15 Tera-FLOPS) serving as the Hong Kong node of the China National Grid (CNGrid), is supported by ITS for shared use with the Department of Computer Science for supporting multi-disciplinary researches and systems research.  The other central HPC server for research support includes a 192-core 64-bit system. 

hpc server

The high throughput computing (HTC) service is particularly useful to problems that are embarrassingly parallel, i.e. those that can be easily divided into smaller, independent sub-problems, each of which can be solved within a reasonable timeframe and with limited resources. The current setup of HTC service comprises a Linux backend server and 235 Windows-based PCs located in the Learning Commons as computer nodes that run HTC jobs only when the Learning Commons is closed, thereby maximizes the utilization of these PC resources, aiming to build a Green IT environment.

For more details on research computing support, please visit