The Euler cluster
SIS operates high-performance computing (HPC) clusters as a central service for ETH Zurich.
The current system, called “Euler”, was introduced in early 2014 and has been expanded regularly since then. The latest “Euler IX” expansion in 2023 brought the cluster’s overall computing capacity to over 200,000 CPU cores and 1,300 GPUs.
Thanks to its open and evolutive architecture, Euler can be used for a wide range of applications, such as HPC, Big Data, Artificial Intelligence and Machine Learning.
In addition to the operation of the underlying hardware, SIS provides a comprehensive software stack on Euler comprising hundreds of scientific libraries, tools and applications, based on open-source as well as commercial software.
This software stack is updated and expanded on a regular basis according to the needs of the cluster’s users. Researchers who develop their own software have also access to multiple programming languages, compilers and tools.
Although Euler is intended primarily for “batch” computations, it also provides interactive tools such as JupyterHub for researchers who are not comfortable with a batch system and/or a command-line interface.
More detailed information, user documentation and application lists are provided in the SciComp Wiki. Euler users who require technical assistance can contact our HPC team at firstname.lastname@example.org.
A small part of Euler is open to all members of ETH and allows them to run small computations (subject to certain CPU and RAM limits) at no cost.
Research groups who need more computing power can become “shareholders” of Euler by financing CPUs or GPUs in the cluster. In return, they get the right to use, on average, a share of the cluster's resources proportional to their investment.
Shareholders are not limited to the resources they bought but can use (far) more resources during peaks, as long as their average usage does not deviate too much from their “fair share” of the cluster. This elasticity is one of the most important features of the shareholder model.
The terms and conditions are described in the corresponding service level agreement. The service owner is Olivier Byrde, head of the HPC Group.
We regularly organizes workshops for beginner and advanced cluster users. In collaboration with software developers, we also organize workshops for particular applications and tools.
Please check the list of our past and upcoming workshops here.