Services for performance
analysis of Linux clusters:
- Linux platform compiler evaluation: Intel,
and PGI
- Linux cluster IA-32 & IA-64 hardware benchmarking.
- Applications benchmarking on Linux clusters.
Some sample performance
analyses from the technical reports pages:
- MPI bandwidth
comparison of the SGI Altix versus new Quad-core nodes
- MPI Parallel
efficiency for CMAQ on a 300+ core cluster
- The Stommel Ocean
Model (in a Fortran 77 MPI version) with the Pentium III SSE instruction set
using the Portland Group Compiler ( for problem sizes 1000 x 1000 to
10,000 x 10,000.
-
For the latest results see the current year's report(s) at
this list for progress with CMAQ and
AERMOD
![](images/servic2.gif)
Table 4.6. Parallel efficiency for the MPI results with CMAQ 4.6.1 using
the EBI ROS3 solvers.
Col x Row = NP |
MPI parallel efficiency |
EBI |
ROS3 |
1 x 1 = 1 |
1.00 |
1.00 |
1 x 2 = 2 |
0.90 |
0.96 |
2 x 2 = 4 |
0.76 |
0.88 |
2 x 4 = 8 |
0.64 |
0.71 |
2 x 8 = 16 |
0.47 |
0.54 |
4 x 4 = 16 |
0.42 |
0.52 |
![wpe3.jpg (38937 bytes)](images/servic0.jpg) |