SCITAS Mission Statement
The main mission of the SCITAS technology platform is to provide researchers at EPFL access to scientific equipment and service expertise in high-performance computing that is efficient and customer-oriented. The subsidiary mission is to contribute, through its own research and development activities, to advancing the technology of the platform for the benefit of its users and the influence of EPFL, and thus to maintain recognized and requested centers of expertise. Finally, SCITAS is involved in teaching HPC and Scientific Computing in Masters and Doctoral School.
The SCITAS platform is used by around 100 labs at EPFL with 800 unique users in 2020. While the majority of users are from computational chemistry, material science, fluid mechanics and engineering, the SCITAS clusters are increasingly used by new communities, such as machine learning, bioinformatics/life sciences or data sciences. Those communities require server configuration and support which are not traditionally provided by HPC platforms. Therefore, SCITAS is currently working on providing services beyond traditional HPC to tackle those challenges, for instance by increasing the size, speed and security of its storage, or providing access to public cloud providers transparently through implementation of a hybrid cloud.
SCITAS engages with its users base, both through its Steering Committee and through regular users meetings, to ensure that it addresses the diverse set of requirements of the different communities that rely on its services.
More precisely, SCITAS focuses on providing:
- resources consolidation and mutualization to reduce local cluster fragmentation and centralize administration in order to minimize financial risks, and reduce the burden of operating and maintaining HPC systems.
- high bandwidth and low latency network, petabyte-scale parallel distributed filesystems shared between the different systems, high performance CPU-only and hybrid CPU/GPU nodes with optimized scientific software stack.
- services such as containers, source repository to promote open science, jupyter notebooks to facilitate access to resources.
- professional operation and maintenance of HPC clusters in order to ensure proper security and data protection level and backup.
- professional HPC software development by HPC experts to maximize user code efficiency on local clusters, but also larger infrastructure such as CSCS or Prace supercomputers.