Odycloud newsletters (2023)

A new GUI to enhance HPC experience in the public cloud

We are pleased to announce the availability of ‘GUI for Numerical Predictions in the public cloud: Graviton3 & 3E.’ A Marketplace AMI with CMAQ and WRF is now available for the 7th generation of AWS EC2 instances powered by Graviton3 with a 5 days free trial. The GUI allows model users to download data, preprocess and run the apps within the same graphical environment. A GIS framework facilitates visualization and enhances user experience. The GUI also has the capability to copy files directly to and from S3 buckets so there is no need for external storage or other on-premises resources if so desired. The AMI is available from https://aws.amazon.com/marketplace/pp/prodview-fcy4llnqba5gk. Bespoke air quality predictions and weather forecasts using near-real time data are also available on request. Weather forecasts The GUI allows to complete all steps necessary to perform WRF simulations from preprocessing of geographical & meteorological data to visualization of the computational results. The preprocessing tools allow model users the completion of several tasks ranging from the download of meteorological files (e.g. GFS, NAM or NRT) in an accelerated fashion to the generation of the domain(s) with the aid of a GIS environment, while also simplifying projection management. Performing the simulations themselves does not require any extra steps as all components are optimized for use in an AWS environment with Graviton 3 processors. The snapshot shows the west-east wind velocity components from a simulation covering 6.25 million square kilometers across parts of Alberta, British Columbia and the Pacific Northwest during the fourth week of August (2023). The simulation used the Lambert conformal conic projection with spatial and temporal resolutions of 7.5 km and 30 seconds, respectively.
Air quality predictions The GUI includes CMAQ v5.4 to perform air quality predictions based on WRF results. To illustrate this capability, the above-mentioned WRF computation was used to model air quality with particular attention to particulate matter concentration caused by the Canadian wildfires. The simulation used a 400 by 275 grid. The below screenshot shows AQI based on PM2.5 (particulate matter with diameters 2.5 micrometers and smaller) on August 20th between 11 am and noon local time (PDT). On this date, the Seattle-Tacoma-Bellevue saw some of the worst AQIs worldwide as captured by the simulation. For example, the measured daily AQIs in Tacoma and downtown Seattle* were 151 and 127, respectively. *Data as reported by Airnow.com for August 20th.

Benchmarks for CMAQv5.4

With the release of CMAQv5.4 at the end of last year, CMAS also introduced a new standard benchmark case 12NE3(2018) partially covering the U.S. eastern region. This benchmark uses a resolution of 12 km similarly to the 12SE1(2016) benchmark, which has been the standard for the last few years. However, the former (12NE3) uses a 105 x 100 horizontal grid slightly larger than the latter (12SE1), which uses a 80 x 100 horizontal grid. The change in domain dimensions renders a direct comparison quantitatively moot but, from a qualitative perspective, the measurements reflect similar trends. We have measured wall times with several AWS 6th and 7th generation compute instances. The figure shows wall times for the 12NE3(2018) benchmark for a period of 24 hours with hourly outputs: A more challenging case is the 12US1 (covering the continental United States) benchmark that has been used with different grids throughout the years. The following results use a 459 x 299 horizontal grid with a 12 km resolution. The figure shows the turnover time for a 24 hours period with hourly outputs:

WRF v4.5

The release of WRF v4.5 by NCAR brought several enhancements to its physics module and a few bug fixes as discussed in the NCAR website (https://github.com/wrf-model/WRF/releases). Our AMIs have updated WRF along with several other elements so that users have access to the latest developments. In order to evaluate performance, we have reassessed the standard 12 km CONUS benchmark. The figure shows wall times for this benchmark using 6th and 7th generation AWS instances.