Flexibility & performance: Google Cloud instead of local infrastructure
CS-21416
HigherEd
University of Vienna
The group "Physics of Functional Materials" is part of the Faculty of Physics at the University of Vienna. The group places a special focus on the use of numerical micromagnetics as well as the development of new algorithms and methods for the computer -aided solution of the micromagnetic equations.
<br>
The Challenge
In times when high computing power is required at short notice, local infrastructures often reach their limits and often the local infrastructure is not even available. Virtual machines do not always offer the desired flexibility in configuration and without the use of containers that would make it possible to install the software according to our ideas.
The Solution
The cloud offers a decisive advantage for us: flexibility. We can quickly and easily test different hardware options. Since we use both CPU and GPU codes, we can exhaust the full potential of the available hardware. This is particularly practical for packages with many dependencies in which a specific version is often required.
The Results
Restrictions that often exist on HPC clusters, such as running time restrictions on jobs, are avoided in the cloud. It is extremely practical to have full access to all cluster settings, which is particularly helpful for debugging. In general, the cloud offers freedom without working unnecessary limits.
The Highlight
The operation of the Google Cloud environment is extremely user -friendly. With the help of the HPC toolkit, a wide variety of applications can be easily implemented in the cloud, be it individual nodes or complex slower clusters. The Google Cloud enables smooth implementation of various requirements.