The Scientific Computing Central Scientific Facility builds and maintains the computing and storage infrastructure for the research in the institute. Apart from supporting the day-to-day computing needs of the researchers, we love to collaborate with them to develop scalable applications, and parallelize and/or optimize algorithms and applications. We also conduct independent research on scalable computing aspects of Intelligent Systems.
Jojumon Kavalan
Our scientists have access to state-of-the-art High Performance Computing facilities. The in-house facilities include a cluster with CPU compute nodes and energy efficient GPU accelerators, and vast amount of storage. Our researchers can also make use of the Max Planck Computing and Data Facility located at Garching.
The in-house cluster features more than 1200 GPU accelerators, including 240 Nvidia H100s. The nodes in the cluster are specially designed for the applications our researchers develop, many having 3TB of RAM and 192 CPU cores each - ideal to run memory intensive scientific programs. The theoretical peak performance of the fast-growing cluster is more than 36 PetaFLOP/s (single precision).
Our scientists work with large amount of data, and we have in-house storage systems with capacity of more than 3 Peta Bytes to address this need. The cluster runs a parallel file system to enable fast data access.
The Max Planck Computing and Data Facility supports our high end computing needs. The supercomputer Viper has close to 100,000 CPU cores and more than 600 APUs, making its peak performance more than 80 PetaFLOP/s (single precision).
We have a motivated team that works with researchers to help them use the computing resources effectively and develop fast and scalable applications.