Visualization: Cerberus: Difference between revisions
Jump to navigation
Jump to search
No edit summary |
(→Basics) |
||
| Line 13: | Line 13: | ||
*** 22TB RAID-6 storage array | *** 22TB RAID-6 storage array | ||
** Network: | ** Network: | ||
*** | *** 1Gbe to public | ||
*** | *** 10GbE to NUC/Jetson cluster | ||
** Coprocessors | ** Coprocessors | ||
*** nVidia Tesla K40c | *** nVidia Tesla K40c | ||
*** 2 x nVidia Tesla K20c | *** 2 x nVidia Tesla K20c | ||
* Software | * Software | ||
** CentOS 7. | ** CentOS 7.9 | ||
** CUDA-10.2 | ** CUDA-10.2 | ||
Revision as of 01:01, 29 July 2021
cerberus.nic.uoregon.edu
Cerberus is one of two publicly reachable login nodes.
Cerberus is the gateway to the "upstairs" mini-cluster, consisting of 16 NUC mini-workstations and 16 Tegra (ARM-64 + CUDA) nodes, and the upstairs provider of large-volume storage (/home/users and /packages) for Cerberus, Chymera, nuc/jetson cluster, xavier, od1k and ryzen.
Basics
- Hardware
- 2 x 10c Xeon E5-2680 v2 @ 2.8Ghz
- 64GB of RAM
- Storage
- 3TB boot disk
- 22TB RAID-6 storage array
- Network:
- 1Gbe to public
- 10GbE to NUC/Jetson cluster
- Coprocessors
- nVidia Tesla K40c
- 2 x nVidia Tesla K20c
- Software
- CentOS 7.9
- CUDA-10.2
OACISS-wide node monitor (Ganglia)
Cerberus runs a Ganglia cluster monitor, accessible from UOnet only (128.223/16) at [[1]].
The ganglia monitor provides continuous real-time updates on the state of all OACISS computing resources.
NUC/Jetson Cluster Manager
Cerberus is the head node for the NUC+Jetson (+Xavier?) mini-cluster.