In May 2021, we have implemented time limit of 10 hours for all interactive shell access to the DGX Kubernetes cluster. This time limit is necessary to avoid anyone from abusing the system by claiming resources but not utilizing them. … Continued
Since we created the DGX cluster, we have been using Kubernetes to run batch jobs. Kubernetes has the basic capability to manage jobs. This is less than ideal since Kubernetes job scheduling is limited. Recently, we have implemented Volcano which … Continued
DGX cluster is getting very busy these days. Jobs are pending for a few days before they get to run. Please use DGX cluster resources carefully. Get what you need to do your computation but not more. Release the resources … Continued
The Kubernetes Dashboard of DGX cluster can be access from the following URL. https://kubem.its.unc.edu:32000/dashboard/
We have completely renovated the VCL instance, TarHeel Linux, CentOS 7 (Full Blade with GPU) to include Podman, Singularity, and Kubernetes. One can use that VCL image to run Docker and Singularity images. Also, one can access and submit jobs … Continued
In October 2018, we add 2 new nodes to the DGX cluster. These 2 nodes are Dell PowerEdge C4140 with 40 physical CPU cores and 256GB of memory each. Also, each of the machines has 4 Nvidia Tesla Volta V100 … Continued
Welcome to dgx.web.unc.edu. This site is created to house information on the DGX GPU Cluster in ITS Research Computing Center.