DEPRECATED
Check Julia 2 documentation instead!
Hardware Overview
Computing Hardware
Julia consists of:
- 52 standard compute nodes
- 7 standard compute nodes with additional infiniband interconnect
- 5 GPU nodes
- 5 prepared GPU Nodes
- 5 FAT Nodes
Standard Compute Nodes
The standard compute nodes are a dual socket setup:
- 2x Intel® Xeon Gold 6134 Processor
- each CPU has 8 cores and can handle 16 threads
- 32 thread in total
- 384 GB of memory
- 10Gbit/s network interface
Infiniband Compute Nodes
The infiniband compute nodes are a dual socket setup:
- 2x Intel® Xeon Gold 6134 Processor
- 384 GB of memory
- 10 Gbit/s network interface
- Additional FDR 56 Gbit/s Mellanox infiniband interconnect
GPU Compute Nodes
The GPU compute nodes are a dual socket setup:
- 2x Intel® Xeon Gold 6134 Processor
- 384 GB of memory
- 10 Gbit/s network interface
- 2x NVIDIA Tesla P100 with 16 GB of memory
prepared GPU Compute Nodes
Like the GPU nodes, but without the GPU cards. The required slots and power supplies are, however, already in place. We can therefore add additional GPU cards if needed.
FAT Compute Nodes
The FAT compute nodes are a quad socket setup:
- 4x Intel® Xeon® Gold 6140 Processor
- each CPU has 18 Cores and can handle 36 threads
- 144 threads in total
- 2 TB of memory
- 10 Gbit/s network interface
Storage
BeeGFS
Julia has 500 TB of storage capacity in a BeeGFS parallel distributed filesystem. It is directly attached to the nodes in the job scheduling system and externally available through julia-storage.uni-wuerzburg.de. Use this storage as the primary place to store your output data from simulations and input data for analysis.
Info
Since we can not provide backup space, the users are responsible for backup of their data.
Ceph
The cloud part is backed by 80 TB of disk space via Ceph for images and volumes.