|
Using WRF in a Cloud Computing Environment
WRF is now supported on Amazon Web Services (AWS) and Scala Computing cloud-computing platforms. The Cloud WRF materials consist of system code and static input data. The supported environments are built with GNU Fortran compilers, along with pre-installed required libraries and WRF binaries (including WPS). See below for specifics on each platform.
WRF on Amazon Web Services
WRF and WPS versions 4.0 to current have been precompiled with a GNU (gfortran) compiler in a pre-configured environment, and saved as an Amazon Machine Image (AMI). For those with an existing AWS account, you may use our public AMI to create your own instance with pre-compiled WRF and WPS code. The AMI includes all libraries necessary to build and run the executables, along with static geographical data (for geogrid), and post-processors RIP and NCL. Please see the README file available in the top-level home directory with additional information.
The public AMI can be found in the US East (N. Virginia) region.
Version 4 code AMI ID: ami-0fd55c94394fa90f9
For a tutorial on accessing and running WPS/WRF code on a pre-configured AWS platform, see this page.
Accessing an AWS cloud environment with a Windows machine
WRF on Scala Computing
Instead of maintaining hardware itself, Scala Computing serves clients through accessing the compute infrastructures of other CSPs. The Scala interface submits jobs to the provider determined optimal at the time, reflecting price and compute request. Users manage their own "projects," which are individual environments configured for their job type. Scala provides configured WRF environments, including installations of the compilers, libraries, WRF and WPS binaries, and static input data. Users running WRF only need to modify their namelists and scheduler scripts and to import meteorological data for each run. This setup is ideal for users repeating consistently configured simulations, such as in a real-time WRF forecasting system.
For the Cloud WRF setup, the Scala Compute Platform provides a development environment currently coupled with an AWS cluster, using a CentOS instance. Scala provides a Network File System (NFS) capability for facilitating simulations and data storage that is mounted on a head node and accessed for the cluster’s instances when a job is submitted. The Scala environment offers sample scripts for submitting WRF jobs, using a Slurm scheduler. Users define their cluster in terms of number of cores, amount of memory, and instance type. For visualization purposes, NCL is included, and the ncview and X11 utilities are installed for quick viewing of model output.
For a tutorial on accessing and running WPS/WRF code on a pre-configured Scala platform, see this page.
Notes
- For detailed information, see the BAMS publication Cloud Computing Efforts for the Weather Research and Forecasting Model.
- The WRF Cloud Computing project is fairly new to our support group. We are working to become more familiar with it and to increase the resources available to our users. More documentation will be released in time. Thank you for your patience.
|