| |
Statistics
Horseshoe5
The Horseshoe5 cluster started production 13/03/2009.
Conceptual overview
- 106x computing nodes (840 CPU cores, 55 TB user disk space)
- 20 GBps fully nonblocking DDR Infiniband interconnect
- Gigabit ethernet
- 1x front-end machine
- 1x infrastructure server
- Theoretical peak performance: 8.5 GFlops
Hardware
Infrastructure
- Front-end Server
- IBM x3550, 2x 2,5Ghz Intel Penrynn 4-core CPUs, 16 GB Ram, 1x500 GB HDD (7200rpm, 8 MB buffer, SATA300), 2x Gigabit Ethernet, 1xDDR infiniband
- Infrastructure Server (batch, queue, web, name-services, mail)
- IBM x3550, 2x 2,5 GHz Intel Penrynn 4-core CPUs, 16GB Ram, 2x500 GB HDD, 2x Gigabit, 1xSDR Infiniband
- Switches
- 3 D-link DGS-3100-48 gigabit switches in a stack
- 1 QLogic Silverstorm 9120 Infiniband switch with 108 20GBps ports
Cluster nodes
- Computing nodes
- 100x IBM x3550 1U rackmountable servers with 2 2,5Ghz Intel Penrynn Low voltage CPUs, 16 GB Ram, 500 GB HDD (7200rpm, 8 MB buffer, SATA300), 2x Gigabit Ethernet, DDR Infinibandx4
- 6x IBM x3550 1U rackmountable servers with 2 2,5Ghz Intel Penrynn Low voltage CPUs, 16 GB Ram, 2x500 GB HDD (7200rpm, 8 MB buffer, SATA300), 2x Gigabit Ethernet, DDR Infinibandx4
Software
- Operating system on nodes and frontend
- CentOS GNU/Linux 5.2 on a 2.6.18 kernel
- OFED 1.3/Infinipath 2.2.1
- Operating system on infrastructure server
- Ubuntu 8.10-server
- Batch- and scheduling software
- Torque 2.3.6
MAUI 3.2.6
- Compilers
- GCC 4.3.2
- Applications and Libraries
- OpenMPI 1.2.5
- See Docs.
|
|
 |
