Computing Infrastructure
To satisfy our diverse compute-intensive needs, we operate a cutting-edge distributed-memory cluster with a high-speed network for small to medium sized jobs.
Our current system, installed in autumn 2019 is a DELL PowerEdge cluster. The system is based on DELL’s PowerEdge R740 platform and consists of 10x regular compute nodes, 2x fat compute nodes, and 1x login/storage node. The nodes are based on INTEL Cascade Lake CPUs. In total, the compute nodes provide 480 physical cores (960 logical cores) and about 3.5TB of main memory.
The cluster uses a 100Gbit/s INTEL Omni-Path high performance network and a 10Gbit/s Ethernet management network. A 48TB high speed, RAID-based storage for temporarily storing simulation data during job execution, is hosted by the login/storage node. NVIDIA Turing architecture based Tesla T4 GPUs provide additional acceleration for specialized workflows, such as machine learning, data analytics, and ray-tracing applications.
Regular compute node (10x)
PowerEdge R740 Server
2x INTEL Xeon Gold 6248, 2.5GHz, 20C/40T
192GB RDIMM, 2933MT/s
480GB SSD
Broadcom 10Gbit/s Base-T
INTEL Omni-Path Host Fabric Adapter
NVIDIA Tesla T4
Fat compute node (2x)
PowerEdge R740 Server
2x INTEL Xeon Gold 6248, 2.5GHz, 20C/40T
768GB RDIMM, 2933MT/s
480GB SSD
Broadcom 10Gbit/s Base-T
INTEL Omni-Path Host Fabric Adapter
Login/storage node
PowerEdge R740XD Server
2x INTEL Xeon Gold 5217, 3.0GHz, 8C/16T
128GB RDIMM, 2666MT/s
24X 2TB SAS
PERC H730P+ RAID Controller
Broadcom 10Gbit/s Base-T
INTEL Omni-Path Host Fabric Adapter
Network switches
S4148T-ON 48x 10GBase-T, 10Gbit/s
H1024-OPF 24x Omni-Path, 100Gbit/s
In addition, we have access to the large-scale high performance computing resources of the Vienna Scientific Cluster.