High-Performance-Computing (HPC)

The Fraunhofer Heinrich Hertz institute runs a High Performance Computing (HPC) Cluster, providing the employees of the departments Wireless Network (WN) and Video Coding and Analytics (VCA) with the necessary computing power to run complex simulations.

The cluster currently consists of nearly 250 computing nodes with about 300 TFLOPs peak performance and with 1.64 PB decentralized storage, as well as 250 TB centralized storage, all linked via a 56 gigabit Ethernet connection. Additionally, the HPC-Cluster contains a high-available virtualization cluster, which provides an environment to run virtual machines as well as other auxiliary resources required for the operation of the HPC-Cluster.

Using so called Workstations the users have access to the cluster resources. The system is totally transparent for the users, this means it looks like it where one.

A liquid cooling solution can dissipate heat very efficiently. The Direct Contact Liquid Cooling- (DCLC-) system used in our cluster is flexible enough to handle varying amounts of heat loss which can fluctuate greatly depending on cluster utilization / load.