 |
|
 |
Simulation of Large Systems: Department
|
|
Information about the mozart cluster
|
|
|
|
|
|
|
Mozart - High Performance Computing Cluster
|
|
General Description
The linux cluster Mozart is a high-performance computing cluster with 64 dual processor nodes and with Infiniband 4x interconnect technology
(see below for details). It is operated by the Chair
Simulation of Large Systems
and the Chair of
Numerics for Supercomputers
and installed at the Institute of Parallel and Distributed Systems
(IPVS) of the
Universität Stuttgart.
It's primary usage will be for teaching and research in the fields of
parallel computing, simulation technology, and simulation applications
such as computational fluid dynamics, molecular dynamics, or traffic
simulations.
For users of this machine a wiki describing howto use it can be found here
Manufacturer
MEGWARE Computer GmbH
Start of operation
March 2004
Fields of application
Parallel computing, simulation technology, numerical simulations.
Technical Details
| Master |
| Motherboard | supermicro X5DPA-GG ATX |
| CPU | 2xCPU Intel Xeon 3,066 FSB533 1MB CACHE |
| RAM | 4 GB DDRAM PC266 ECC |
| HDD | IBM Ultrastar 146Z10 73GB 68pin |
| OS | Linux RedHat 9 |
| gcc version | 3.2.2 |
| 64 Nodes |
| Motherboard | supermicro X5DPA-GG ATX |
| CPU | 2xCPU Intel Xeon 3,066 FSB533 1MB CACHE |
| RAM | 4 GB DDRAM PC266 ECC |
| HDD | IBM 180GXP/60GB 7200rpm |
| OS | Linux RedHat 9 |
| gcc version | 3.2.2 |
| Networking |
| Administrating | 1Gb/s Ethernet |
| Computing | 8Gb/s Infiniband |
HPL Benchmark
MPICH2 is the MPI implementation installed on Mozart, with
Infiniband support.
Results of the HPL (High Performance Linpack) benchmark
![[Performance vs. Processors]](pictures/perf_proc.jpg) |
Figure 1. The performance scales up with the number of processors.
|
|
| |
![[Problem size vs. Processors]](pictures/pbsize_proc.jpg) |
Figure 2. The problem size scales up with the number of processors.
|
|
| |
![[Performance vs. problem size]](pictures/perf_pbsize.jpg) |
Figure 3. The performance scales up with the problem size.
|
|
| |
![[Percentage]](pictures/percentage.jpg) |
Figure 4. The evolution of percentage of the peak performance achieved with the increasing of number of processors.
|
|
|
Mozart Gallery
 |
| Front view of Cluster Mozart |
|
 |
| Back view of Cluster Mozart |
|
 |
| Life can be complicated |
|
 |
| The Master |
|
|
|
|
|
|
N.N. N.N.
)
| © Universität Stuttgart |
Legal notice