ITSC provide high-performance computing (HPC) service to HKUST researchers. The HPC service in HKUST is organized base on the Purdue community cluster model, which is widely adopted by research universities all over the world. The HKUST research community cluster solicit Schools or Departments to contribute funding for procurement of high performance computing machines, while ITSC would assist in coordinating joint purchase, provides the necessary infrastructure, managed data center hosting service, cluster system administration services and hardware maintenance of the machines. Contributors do not solely own the hardware resources while the machines are in the community cluster, instead they have priority to use their contributed resources. Whenever there are idle resources, these spare resources would be made available to other users of the HPC cluster to maximize the overall resource utilization. The machines in the HPC cluster runs with a lifespan of five years and would be subjected to the retirement review from the 6th year onward to redistribute the scarce resources of data center space to new contirbutor when needed.
In response to the substantial demand for HPC computing resources at HKUST and to optimize the utilization of the limited resources available within the HPC data center for the broader research community, the ITSC can solely accommodate HPC machines that are willing to integrate into the community cluster.
As the decommissioning process of HPC2 progresses, the ITSC presently upholds one operational high-performance computing cluster, ensuring continued accessibility for the university community. Additionally, there is ongoing planning and development of a new high-performance computing cluster, the HPC4, slated for production use in the second quarter of 2024.
The HPC3 cluster, established in Spring 2020, is a contributed cluster by faculty members which is designed to cater both CPU-intensive computational jobs and jobs requiring GPU coprocessor support. It is open to all approved university researchers but cluster contributors have higher access priority.