Hardware

   1. Short hardware news
   2. Current hardware: main characteristics
   3. Current hardware: comparison
   4. Current hardware: system load
   5. Technical support
   6. Decommissioned hardware

1. Short hardware news

• 01/2012: The Cray system hexagon will be upgraded from an XT4 to an XE6 in March 2012. The new system will have 696 nodes (each with two AMD Interlagos 16-core CPUs, 2.3 Ghz), a total of 23040 cores, and includes the Gemini interconnect. The upgraded system will be available for Norwegian research as of period 2012.1 (start April 1 2012). Projects can apply for access to hexagon for the period 2012.1 using the regular procedures for resource allocation in Notur.
• 01/2012: A new HPC system vilje will be available for Norwegian research as of period 2012.1 (start April 1 2012). Projects can apply for access for period 2012.1 using the regular procedures for resource allocation in Notur. A test system ve is available for early porting of applications. Contact sigma@uninett.no if you would like to have access to that system. Vilje replaces njord that will no longer be available for general access after February 1.
• 12/2011: A new Nordic HPC system gardar will be available for Norwegian research as of January 1, 2012. In the first three months of its operation, it will be made available to selected projects. Projects can apply for access for period 2012.1 (start April 1, 2012) using the regular procedures for resource allocation in Notur.
• 12/2011: The facility njord will be taken out of operation in 31 December 2011. Only for special purposes, access to some nodes of njord may be possible after December 31.

2. Current hardware: main characteristics

The university partners of the Notur consortium operate a number of High Performance Computing facilities. This section lists some characteristics of these facilities.

Each of the facilities consists of a compute resource (a number of compute nodes each with a number of processors and internal shared-memory, plus an interconnect that connects the nodes), a central storage resource that is accessible by all the nodes, and a secondary storage resource for back-up (and in few cases also for archiving). All facilities use variants of the UNIX operating system (Linux, AIX, etc.). Click on the name of the facility in the table below to get further information for that facility.

In case these pages do not provide the information that you are looking for, contact the local support staff, especially if you consider to apply for access to one or more of the systems.

3. Current hardware: comparison

The table below attempts to compare the available systems with respect to the type of applications they are suited for.

• 'yes': the system in question is suitable for the type of jobs/applications
• 'no': the type of jobs on the system is impossible or discouraged
• A number in the table specifies the maximum number of processor cores that can be used (a physical limit or practical experience)

The following considerations should be kept in mind when selecting a system to execute applications:

• The number of cores that can be used simultaneously for shared-memory (OpenMP) applications is limited by the number of processor cores in a compute node of the system. This number is given in the column OpenMP applications.
• Hexagon is a system for large scale parallel (distributed-memory) applications. Applications that use less than 128 cores (or 32 nodes) on hexagon are discouraged. Requests for access to hexagon to execute applications that use fewer cores are often rejected or moved to other systems. Hexagon is used for various applications in climate modeling and computational physics.
• Stallo and titan are throughput systems. These systems can be used for sequential (single-threaded) as well as parallel applications. The systems are used for most of the bioinformatics and chemistry applications.
• Gardar, hexagon, stallo and titan have standard Linux distributions as operating system. Njord uses IBM's AIX operating system and IBM's proprietary compilers and libraries. In case you need to install a specific software package, please make sure that you know for which environments the software is supported, before choosing a system.
• See also the list of application software that is available on the Notur systems.

In case these pages do not provide the information that you are looking for, contact the support staff, especially if you consider to apply for access to one or more of the systems.

In case you have a specific software application that you wish to execute on the Notur systems, but you are unsure which system is most suitable, we strongly advise you to contact the support staff.

4. Current hardware: system load

A separate page gathers the live status (load) for each of the Notur facilities.

5. Technical support

On the support pages you find more information on how to contact the support staff and where to find the technical support pages for the facilities in the Notur project.

6. Decommissioned hardware

• The HP Itanium2 cluster snowstorm was taken out of Notur on March 31, 2008
• The HP SuperDome magnum was taken out of Notur on December 31, 2007
• The IBM p690 tre was taken out of Notur on September 30, 2007
• The IBM 1300 cluster fire was taken out of Notur on March 31, 2007
• The SGI 3800 gridur/embla was taken out of Notur on December 31, 2006
• The SGI Altix snehvit was taken out of Notur on September 30, 2006
• The HP SuperDome nana was taken out of Notur on March 31, 2005.