Technical Specifications of NIWA’s HPCF

Science Centres: Atmosphere, Climate, Coasts, Energy, Natural Hazards, Oceans

The specifications for our current machine, and those of the system after it is upgraded in 2011/2012.

Phase 1 (2010)

50170

Hardware

  • IBM p575/p6 supercomputer with 56 POWER6, 32 way 4.7 GHz nodes for a total of 1792 processors and 5.4 terabytes of memory;
  • Capable of performing up to 34 trillion (million million) calculations per second
  • The bandwidth within each 32 way node is as high as 80 GB/s. The nodes can communicate with each other at 16 GB/s per connection in each direction with a latency of around 4μs (MPI) for short messages.
  • 0.5 petabytes of useable disk storage on 740 disks in an IBM DCS9900 storage array;
  • Two IBM TS3500 Automatic Tape Libraries, with 12 LTO-5 drives and each capable of storing 2.5 Petabytes of data. One tape library will be for disaster recovery, and will contain copies of all data on the primary tape library;
  • Eight p520/p6 servers for HPC Management, GPFS (General Parallel File System), and TSM (Tivoli Storage Manager) functions; and
  • One BladeCenter with 96 Xeon 2.53 GHz processors and 384 Gigabytes of memory for pre- and post- processing tasks.
50171

HPC Software Stack

  • AIX operating system on all p520 servers and p575 nodes;
  • General Parallel File System (GPFS) – accessible from all HPCF processors – both POWER 6 and Bladecenter X-Series;
  • Tivoli Storage Manager (TSM) to provide backup and transparent movement of data between storage media (Hierarchical Storage Management);
  • Tivoli Workload Scheduler software – to manage and schedule work on the HPCF;
  • XL Fortran, C and C++ compilers, and
  • IBM High Performance Computing Toolkit.
50172

Physical Infrastructure

  • 370 kW to run the system;
  • Another 120 kW for cooling, using chilled water (which removes around 70% of the heat) and air (which removes the rest of the heat);
  • Weight: 12.6 tonnes, including water.


Phase 2 (2011/12)

50173

The new computer hall has been built to support a planned upgrade to

  • An additional 52 POWER6, 32 way 4.7 GHz nodes, for a total of 3456 processors and 9.0 terabytes of memory;
  • Be capable of performing 65 TeraFLOPS;
  • An additional 1.4 Petabytes of disk storage for a total 2 Petabytes (i.e. two million gigabytes)

The existing computer hall has already been designed to

  • supply 675 KW to run the expanded system
  • supply the additional 225 KW to cool the upgraded system
  • Support the final weight of 18.8 tonnes, including water