NJORD
| |
System |
Type |
Number of nodes |
Number of cores |
CPU type |
Theoretical total peak |
Total memory |
Total disk capacity |
| njord |
IBM p575+ |
distributed SMP |
65 |
944 |
Power5+ |
7.2 Tflop |
2272 GB |
120 TB |
Full name: njord.hpc.ntnu.no
A summary is provided below.
Detailed information can be found on Notur's technical support pages.
Njord is a distributed shared-memory system. Its name comes from the Norwegian god Njord who is the protector of seafarers and fishermen; he sends favourable winds and calm seas. The system consists of IBM p575+ nodes interconnected with a high-bandwidth low-latency switch network (HPS).
The system has a total of 65 nodes partitioned into 59 compute nodes, 4 I/O nodes and 2 login nodes:
- all 59 compute nodes are shared memory nodes with 8 dual-core power5+ processors 1.9 Ghz
- 55 compute nodes have 32 GB memory, four compute nodes have 128 GB memory.
The disk storage system is accessed through IBM's distributed parallel I/O system GPFS (General Parallel File System).
The system is well-suited for large scale parallel MPI and OpenMP applications, as well as applications that combine these two communication paradigms.
Other details:
Operating system |
AIX 5.3 |
Queueing/scheduling system |
LoadLeveler |
Total space for shared temporary storage |
33 TB |
Total space for home directories |
48 TB |
Max addressable memory |
55x13 + 4x128 GB |
| |
| User limits |
Default size home directory |
20 GB |
| |
| Job limits |
Maximum job run-time |
21 days (504 h) |
Maximum number of CPUs per job |
55 nodes (1760 tasks) |
Location: NTNU, Trondheim
Installation: September 2006
Included in Notur: December 2006
| 