MSU Denver Brings AWIPS II to the Classroom

MSU Denver

Editor's Note: Metropolitan State University of Denver (MSU) is one of a handful of universities of varying sizes that have incorporated AWIPS II into labs and classrooms in the past two years. MSU's experience shows that incorporating the new NWS technology into the educational mix is both possible and useful even for a small undergraduate program. This article outlines the MSU AWIPS II configuration and discusses some of the adjustments the program made to get the system running smoothly.

Sam Ng at MSU
Professor Ng working with students in the Weather Analysis Techniques lab.

Thanks to a Unidata Community Equipment Award grant coupled with some local resources from our university, we (Professor Sam Ng and co-PI Chris Kimmett) at Metropolitan State University in Denver have been able to successfully bring the National Weather Service's AWIPS II forecasting system into the classroom. Professor Ng used AWIPS II alongside Unidata's Integrated Data Viewer as core tools in the Spring 2015 “Weather Analysis Techniques” course for undergraduate meteorology majors, and will continue using AWIPS II in the Fall semester and into the future.

Hardware Used

The AWIPS II hardware configuration specified by the NWS is designed for operational use in Weather Forecast Offices (WFOs). As such, it consists of several high-powered server machines making up the Environmental Data EXchange (EDEX), which ingests, processes, and serves real-time data to client computers running the Common AWIPS Visualization Environment (CAVE) software for display and analysis. Developers at the Unidata Program Center designed a slimmed-down EDEX server configuration consisting of a single server computer, which is more suitable for a non-operational setting such as a university classroom. MSU's EDEX server is configured as follows:

  • HP ProLiant DL380p Gen 8
  • 2 x 16 core Intel Xeon E5-2650 v2 @ 2.6 GHz
  • 32 GB RAM ( 8 x 4 GB 1866 Mhz DDR3 SDRAM )
  • 146 GB 15K HDD - Operating System
  • 1.8 TB 10K HDD - Storage
  • 300 GB SSD - Radar Data
  • 1 Gigabit Network Interface
  • RHEL 64 bit Server - version 6.6

This server, which cost roughly $11,000 to configure in early 2015, has proved capable ingesting all of the real-time data we desire (including NEXRAD3 Radar, GEFS, and 0.25° GFS model output) and supporting 20+ CAVE clients simultaneously in our weather lab.

In preparing our weather lab to take advantage of AWIPS II, we upgraded 27 existing workstations with extra memory and more capable video cards, giving us the following configuration:

AWIPS II CAVE and the IDV running on a lab machine.
  • Dell Optiplex 960
  • Intel Core 2 Duo CPU @ 3.0 GHz
  • 8 GB RAM (800MHz DDR2 SDRAM)
  • 160 GB 7.2K HDD
  • 1 Gigabit Network Interface
  • Nvidia 2GB PCIe x16 graphic card
  • RHEL 64 bit Workstation - version 6.6

These lab workstations have proved capable of running both the AWIPS II CAVE client and Unidata's IDV simultaneously.

AWIPS II System Configuration

We began by installing the AWIPS II version 14.2.1 package created by Unidata. Installation itself was relatively straightforward, but getting AWIPS II running without crashing was a challenge. We found that several configuration files needed adjustment from the defaults to run smoothly in our environment.

First, we had to adjust the default pqact.conf file, which controls the actions of the Unidata Local Data Manager (LDM) software that is used by AWIPS II for data ingest. Ingesting model output for the Global Environmental Multiscale Model (GEM) from the Canadian Meteorological Centre (CMC) and the NAVy Global Environmental Model (NAVGEM) caused AWIPS II's GRIB file decoder to crash, due to errors in the GRIB tables. Removing these models (which we were not using in the classroom anyway) from the LDM configuration was as simple as commenting out the corresponding lines in our pqact.conf file, and solved the crashing problem.

Second, we adjusted the HEAP size request in the and wrapper.conf files to fix an issue we experienced when opening multiple CAVE clients. We found that increasing the MAX_MEM parameter in /awips2/edex/etc/ to 4096 resolved a heap memory issue that would crash the ingestGrib process. We also found that changing the parameter in /awips/qpid/conf/wrapper.conf to 4096 resolved a heap memory issue the would crash the qpid process.

After making these adjustments, we have found the AWIPS II 14.2.1 system to be quite stable, with CAVE clients running for weeks at a time without rebooting or crashing. (Most recently, we haven't had a crash since the beginning of the Fall term in mid-August 2015.) We are indebted to Unidata Program Center staffers Michael James and Jeff Weber for helping us diagnose these issues and devise the appropriate solutions.

Note that our initial experiments with the AWIPS II 14.4.1 release in the Fall 2015 have not been as successful, because version 14.4.1 of the CAVE client software does not work with EDEX servers running version 14.2.1.

System Performance

Once the adjustments to our EDEX configuration were in place, the system has had no trouble ingesting the data we are requesting. Our CAVE client machines have similarly been up to the task of allowing up to 20 students at a time to access the system. We have found AWIPS II to be well suited for nowcasting and forecasting exercises in our Weather Analysis Techniques and Mesoscale Meteorology courses.

An Alternative Workflow

AWIPS II on a Mac
AWIPS II CAVE displayed on a Macintosh

The AWIPS II software environment runs only on Linux servers and workstations. While this is not an issue for the dedicated EDEX server or the lab workstations, this limitation was a bit of a stumbling block for Professor Ng, who uses a Macbook Pro to run classes and in daily use. As it turns out, it is possible to install the CAVE client software on a Macbook by running a Linux virtual machine -- in this case using the VMware Fusion software to create a virtual RedHat Linux workstation. Given sufficient network bandwidth, the CAVE client performs with no noticeable delays. In MSU's case, the network is fast enough even over a WiFi connection when using the campus Virtual Private Network (VPN). Your mileage may vary...


Post a Comment:
Comments are closed for this entry.
News and information from the Unidata Program Center
News and information from the Unidata Program Center



Developers’ blog

Recent Entries:
Take a poll!

What if we had an ongoing user poll in here?

Browse By Topic
Browse by Topic
« September 2020