You are here
Time: Fast computing
What is HPC and why should I use it?
Many researchers use computers but desktop machines only go so far. If your overnight compute jobs run into the next day, if your research waits for a weekend to run, if your computer is limiting the progress of your research, then high performance computing (HPC) is the solution.
High performance computing is used to solve real-world problems of significant scale or detail across a diverse range of disciplines including physics, biology, chemistry, geosciences, climate sciences, engineering and many others.
Please email Time@intersect.org.au for questions, advice and support.
Intersect’s HPC facilities
Intersect has a partner share in the peak facilities at the National Computational Infrastructure, based at the Australian National University. In addition to this, Intersect also manages a state facility, Orange hosted at IC2 in Sydney.
Intersect's new HPC 'Orange' was commissioned in March 2013. The SGI 30+ TFlop distributed memory cluster provides a greater than 25-fold increase of compute power and a fivefold increase of disk capacity on the existing system. The new system features 100 cluster nodes with 1600 cores powered by the Intel® Xeon® E5-2600 processor series. It also includes 200 TB local scratch disk space, 101TB of usable shared storage delivering 30+ TFlops, more details here
NCI's new peak system 'Raijin' is a Fujitsu Primergy high-performance, distributed-memory, cluster based on Intel Sandy Bridge 8-core processors (2.6 GHz) comprising:
- 57,472 cores in the compute nodes;
- approximately 160 TBytes of main memory;
- Infiniband FDR interconnect; and
- approximately 10 PBytes of usable fast file system (for short-term scratch space).
This provides a:
- peak performance of approximately 1.2 PFlops -- 8.6 times that of the current peak system;
- sustained performance of approximately 6.7 times the current peak system (i.e., an aggregate SPECfp_rate2006 of 1.6M); and
- file system performance which is 6.0 times that of the current system; and consistent with NCI's requirements that the system be well-balanced.
The unit of shared memory parallelism is the node, which comprises dual 8-core processors, i.e., 16 cores. The memory specification across the nodes is heterogeneous in order to provide a configuration capable of accommodating the requirements of most applications, and providing also for large-memory jobs. Accordingly:
- two-thirds of the nodes have 32 GBytes, i.e., 2 GBytes/core;
- almost one-third of the nodes have 64 GBytes, i.e., 4 GBytes/core; while
- two (2) per cent of the nodes have 128 GBytes, i.e., 8 GBytes.core.
If you use resources on Orange or Raijin via the Intersect partner share we ask that you acknowledge us. The proposed text is:
Computational (and/or storage) resources used in this work were provided by Intersect Australia Ltd.
The full policy can be found here: http://www.intersect.org.au/attribution-policy