HPC Facilities
NeSI provides High Performance Computing (HPC) facilities, supporting researchers in New Zealand across the public research sector and in private industry. As a national High Performance Computing infrastructure, we support researchers with state-of-the art integrated systems, facilities and infrastructure, while providing expertise and support.
NeSI supports HPC systems at three facilities, being The University of Auckland, Canterbury University, and NIWA. The groups supporting HPC at each site are the Centre for eResearch, BlueFern, and FitzRoy High Performance Computing Facility respectively.
| Platform | Hardware | Operating Environment | Facility |
|---|---|---|---|
BlueGene/P | 2048 4-core CPUs (8196 0.8 GHz processor cores) 8192 GB RAM, 1 GB RAM per core 3 Dimensional Torus interconnect between CPUs | Linux GPFS high performance parallel file system LoadLeveller workload scheduler and TSM-SM Hierarchical Storage Management (HSM) | |
P575/POWER6 | 58 nodes each with 32 cores (1856 4.7 GHz processor cores) with 64 and 128 GB RAM nodes total 5376 GB RAM increasing to 108 nodes (3456 processor cores) total 8704 GB RAM from 2012 DDR Infiniband interconnect between nodes | AIX GPFS high performance parallel file system LoadLeveller workload scheduler and TSM-SM Hierarchical Storage Management (HSM) | |
P755/POWER7 | 13 nodes, each with four 8-core CPUs (416 3.3 GHz processor cores) 128 GB RAM per node, total 1664 GB RAM Infiniband interconnect between nodes | Linux, AIX GPFS high performance parallel file system LoadLeveller workload scheduler and TSM-SM Hierarchical Storage Management (HSM) | |
Intel cluster | 76 nodes, each with 11 available cores (2.68 Ghz) and 96 GB RAM per node and 124 nodes, each with 15 available cores (2.68 GHz) and 128GB RAM 12 nodes each have 2 x Tesla M2090 GPUs QDR Infiniband interconnect | Linux GPFS high performance parallel file system LoadLeveller workload scheduler and TSM-SM Hierarchical Storage Management (HSM) | |
Intel Visualisation cluster | 5 iDataplex nodes, each with dual 4-core 3.03 GHz Intel Xeon CPUs 2 Tesla M2070Q GPE per node 96 GB RAM memory per node | Linux GPFS high performance parallel file system TSM-SM Hierarchical Storage Management (HSM) |
Scientific applications
Facilities install and maintain a wide variety of computational science applications and packages to support your research, though only for applications suitable to be run on HPC facilities. Licences are available for some application packages, while in other cases licences are required to be provided by the researcher.
Capability and Capacity Computing
Capability computing established a balance of computing resources, across memory, CPU cores, process interconnect, and data network, to provide a well tuned and matched optimum performance for a certain class of computational workload. Often such a system can solve computational problems far faster than any system without similar balance.
Capacity computing is far less optimised for specific classes of workload, rather it employs cost effective commodity componentry to deliver more general-purpose systems. These general-purpose systems are sometimes tuned and configured for larger workloads, and often also simply used for many tasks, or to aid in preparation for use of capability computing.
Fit-for-Purpose
A measure of Fit-for-Purpose is sometimes applied when assessing the match between computational workload and computing system, especially when looking to make optimal use of capability computing systems. With the expense required to create capability systems, this approach is applied in order that the most suitable workloads make efficient use of the balance of such an expensive research instrument.
Allocation of HPC facilities
NeSI provides an equal opportunity merit based allocation mechanism for any researcher working at a New Zealand public sector research institute or in private industry.
