Easy Deployment of the WRF Model on Heterogeneous PC Systems Braden Ward and Shing Yoh Union, New Jersey
Easy Deployment of the WRF Model on Heterogeneous PC Systems
Jan 01, 2016
Easy Deployment of the WRF Model on Heterogeneous PC Systems. Braden Ward and Shing Yoh. Union, New Jersey. Bootable Cluster CD (BCCD). Features 2.4 Linux Kernel Available online http://bccd.cs.uni.edu/ - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Easy Deployment of the WRF Model on Heterogeneous PC Systems
Braden Ward and Shing Yoh
Union, New Jersey
Bootable Cluster CD (BCCD)
Features 2.4 Linux Kernel Available online http://bccd.cs.uni.edu/ Networked file system used for data storage Setup time of 1-2 minutes per PC OS and file system completely undisturbed
upon restart
Using the BCCD
3 Groups of PCs Used
448 MHz Pentium 3 desktops with 256 MB of memory (networked 100 Mbps switches)
2.8 GHz Pentium 4 desktops with 2.0 GB of memory (networked 100 Mbps switches)
16 node Linux cluster: 3.0 GHz Pentium 4 processors, 1.0 GB of memory each, linked with 1 Gbps switches
Supercell Ideal Case
em_quarter_ss in WRF distribution 2 hours forecast Grid resolution of 2 km (100 x 100 x 40)
0
10
20
30
40
50
60
70
80
0 4 8 12 16
Pentium 3Pentium 416-node Cluster
Supercell Ideal CaseR
un
tim
e (m
inu
tes)
Number of Processors
Real case
12 hour forecast, beginning 00UTC February 11th, 2006
Grid resolution of 15km (100x100x31 grid points)
Initial and boundary conditions generated from 00Z NAM run from NCEP using WRFSI
0
20
40
60
80
100
120
140
160
0 2 4 6 8 10 12 14 16 18
Pentium 3Pentium 416-node Cluster
Real CaseR
un
tim
e (M
inu
tes)
Number of Processors
0
10
20
30
40
50
60
70
80
0 4 8 12 16
Pentium 3Pentium 416-node Cluster
Supercell Ideal CaseR
un
tim
e (m
inu
tes)
Number of Processors
0
20
40
60
80
100
120
140
160
0 2 4 6 8 10 12 14 16 18
Pentium 3Pentium 416-node Cluster
Real CaseR
un
tim
e (M
inu
tes)
Number of Processors
Results
A Linux cluster is easily created using PCs with BCCD
Real case runtimes decreased 2-4 times on clusters of 4-8 PCs
Useful for students and operational forecasters with limited or dated resources to run a locally configured WRF Model
Acknowledgment
Dr. George Chang from the Computer Science Department and Computational Science Group at Kean University for building the Linux Cluster and providing technical support.
Dr. Paul Gray of the University of Northern Iowa for developing the BCCD and making the software freely available.
The project is supported by the Kean University Office of Research and Sponsored Programs and the Department of Geology of Meteorology.
Related Documents
