|Début du projet||26.03.2018|
|Fin du projet||30.06.2022|
|Thèmes||Système de mesure et de prévision|
The project Framework goal is to bring the software to produce weather forecasts and support forecaster in their duties to a modern IT-architecture. We will use a container platform to support a DevOps philosophy and automatize further the deployment of applications.
To support weather forecasting by forecasters and to create various intermediate and end-user products, several 100 tools and scripts are used in the MeteoSwiss’ forecasting offices. They run alongside main tools that provide data for the MeteoSwiss app and website. The development of these tools partly goes back 20 years. It is a mixture of externally and internally developed applications, which process weather data from different databases, receive inputs from experts, and process them further.
The resulting application architecture could be largely improved by:
- Simplifying application monitoring
- Removing duplicated piece of code
- Implementing systematically DevOps practice and automation to reduce maintenance cost.
We decided to move to service based architecture and implement it on a container platform. We started of with two central applications.
The Data4Web application main task is to deliver data to the MeteoSwiss app and website. All local forecasts are based on these. Below an example of the visualization on the MeteoSwiss.
The actual Version of Data4Web is a mix of different languages, developed by many different developers, over many years. It has become a central piece of the MeteoSwiss production system, as its data are consulted about a million times per day.
We chose to redesign completely Data4Web to reach modern IT standards on security, testing, deployment and monitoring. The application will be split in a suite of webservices, which will run on a Docker based container platform.
Forecast Production Editor and Product manager
The manual production in the three forecasting centers in Geneva, Locarno and Zurich are scheduled by a Production Manager, a kind of daily todo list for each forecaster in shift. It is integrated with a product editor, which is used by the forecasters to issue written products such as the daily weather bulletin below.
Both theses software are based on old technologies and must be brought to actual IT-standards. Further, it will allow us to better integrate them in our new container-based architecture.