DHPCDelft High Performance Computing Cluster


Prof. Kees Vuik
Mekelweg 5, 2628 CD Delft

The TU Delft High Performance Computing Cluster (DHPC) is a heterogeneous computing cluster, consisting of the newest computing and data nodes, connected by a high speed communication network based on a huge amount of memory. DHPC will be embedded in a complete computing ecosystem consisting of ICT help desk and courses given by ICT and DCSE (TU Delft Institute of Computational Science and Engineering) in order to use the DHPC in an optimal way. Due to the large increase of data, it is required to bring compute nodes to the data instead of data to the compute nodes. The transport of data takes much more time than the computations. Combining the modeling, algorithmic, computer science capabilities with the newest hardware, the infrastructure facilitates scientific breakthroughs in a number of fields such as better predictability of Alzheimer, better treatments of burn wounds (less use of test animals), new tools for finance, and robust Quantum Computers.

The DHPC consists of a high number of computing nodes, high memory nodes and GPU nodes, storage and a high speed interconnect. Besides the Computing facilities, the infrastructure comprises an HPC ecosystem where users, students and trainers work together in order to obtain the best results in High Performance Computing for Science and Engineering. The ecosystem relies on a lab room (Penquin lab) where MPI (Message Passing Interface), GPU (Graphics Processing Unit), OpenFOAM (Open source Field Operation And Manipulation) and other courses for both students, staff and non-academic users are given. The DHPC supports computational variants of all sciences, fields such as Computational Fluid Dynamics, Computational Mechanics, Computational Finance (FinTech), Computational Social Science, Machine Learning, Deep Learning and many more.

In computational science two trends are visible: one is the tremendous increase in the amount of data (measurements/simulations) needed in science (AI, deep learning, digital twin, etc) and the second is the fact that the increase in computational speed is much faster than the increase in data transport capabilities. For this reason it is important to have the compute facilities on campus, close to the data. Since the TU Delft is home of the largest group of Computational Science and Engineering researchers (of excellent quality) in the Netherlands it is clear that the best place for a High Performance Computing Cluster is at the TU Delft.

Aansluiting bij strategische ontwikkelingen
High Tech Systemen en Materialen
Waardecreatie door verantwoorde toegang tot en gebruik van big data
Smart industry
Kwaliteit van de omgeving