White Papers

21 Dell HPC Omni-Path Fabric: Supported Architecture and Application Study June 2016
Figure 11 OSU Bandwidth values based on Intel
®
Xeon
®
CPU E5-2697 v4 processor.
5.3 Weather Research Forecast
Weather Research and Forecasting Mode
l
[8] is a weather prediction system designed for atmospheric
research and operational forecasting needs. The WRF system contains two dynamic solvers, referred to as
the ARW (Advanced Research WRF) core and NMM (Nonhydrostatic Mesoscale Model) core. The WRF-ARW
code was used for testing.
Figure 12 illustrates the application performance for node counts scaling from 1 to 32. All the data points in
the graph are relative to one node. As the node count increases most of the MPI time is spent on MPI_Bcast
for this dataset. This application scales linearly across all the nodes counts. Higher values represents better
performance and at 32 nodes, a 44.2X improvement is seen over a single node.
Figure 12 Weather Research Forecasting model performance graph
4
6.4
8.7
10.1
14
16.9
21.2
22.5
23
23.9
24.5
24.3 24.3
3.3
5.1
7.1
7.8
10.2
11.8
12.0
12.1
12.3
12.3 12.3
12.3
12.3
0
5
10
15
20
25
30
Bandwidth(GB/s)
Message Size(bytes)
OSU Bandwidth
Bi-directional Uni-directional
1
2.1
4.1
9.2
20.7
44.2
0
10
20
30
40
50
1 2 4 8 16 32
Relative to one node
No. of Nodes
WRF - Conus 2.5KM
Scaling