CONTENTS
General
Fall 2014
PREVIOUS NEXT
Updates on Research Computing Facilities
Charles Choy
Head of Academic Computing
High Utilization of High Performance Computing Facilities

Equipment Upgrade

Our two high performance computational clusters, namely HPC1 and GPU, were installed 2 years ago and their utilization has come close to saturation. As such, a partial upgrade of hardware equipment was carried out this summer:

HPC1 cluster –

Upgrade to increase computational capacity and CPU performance, and add GPU support, namely:

  • 12 x IBM NextScale DX360 M4 servers with Dual Xeon E5-2670v2 10 cores, 64GB RAM
  • 4 x HP SL250 servers with Dual Xeon E5-2660 v2 10 cores, 64GB RAM, dual Nvidia Tesla K20 GPU

GPU cluster –

Upgrade to increase computational capacity and CPU performance, add new generation of GPU and Xeon Phi coprocessors to support more variety of computational needs, add InfiniBand high speed network support and increase disk storage, namely:

  • 6 x HP SL250 servers with Dual Xeon E5-2660 v2 10 cores, 64GB RAM, dual Nvidia Tesla GPU
  • 6 x HP SL250 servers with Dual Xeon E5-2660 v2 10 cores, 64GB RAM, dual Intel Xeon Phi
  • 2 x HP SL4540 servers with total of 360TB raw storage
  • Both old and new servers connected to 40Gbps InfiniBand network
Furthermore, the GPU cluster will be managed by the SLURM queuing system to improve resource allocation effectiveness.

The new equipment will be on production from Fall 2014. More details are available at:

http://itsc.ust.hk/services/academic-teaching-support/research-computing

Upcoming Initiatives

ITSC is currently working with the Taskforce for Data Science Initiative led by Prof I-Ming Hsing and existing users of high performance computing to identify a sustainable arrangement for supporting the computational needs of academic activities pertaining to Data Science. Initial ideas include:
  • Promoting the Community Cluster Model for effective acquisition and sharing of high performance computing resources

  • Enhancing the existing clusters to meet new needs arising from big data research (e.g. scalable storage infrastructure, Hadoop computing environment, etc.)

  • Facilitating the acquisition and sharing of common software required by researchers
HPC1 Cluster
GPU Cluster