The August update includes the addition of new compute nodes in the ub-hpc cluster, an update to Open OnDemand, new software available, and information about the transition to the new Coldfront portal.  See below for more details.



We are happy to announce the addition of new compute nodes acquired as part of a $1 million grant from the National Science Foundation.  More details about the award can be found here


As of today, CCR has added 118 new compute nodes to the ub-hpc cluster.  These nodes include general computational nodes, larger memory nodes, and GPU nodes.  Below are the specifications for the nodes.  Users can continue to access the large memory nodes using the largemem partition and the GPU nodes can be accessed by selecting the gpu partition.  The new general compute nodes are in the skylake partition.

---------------

Partition: skylake

---------------

total nodes => 86

gpu_count => 0

cores_per_node => 32

memory size => 187.28 GB

product name => Dell PowerEdge R440

SLURM CPU constraint => --constraint=CPU-Gold-6130

SLURM mem setting => --mem=187000


---------------

Partition: gpu

---------------

total nodes => 16

gpu_count => 2

cores_per_node => 32

memory size => 187.41 GB

product name => Dell PowerEdge R740

SLURM CPU constraint => --constraint=CPU-Gold-6130

SLURM mem setting => --mem=187000

SLURM gres setting => --gres=gpu:V100:ngpu  (where ngpu is the # of desired GPUs)


---------------

Partition: largemem

*NOTE: this partition also includes older nodes with relatively large amounts of memory.  The new largemem nodes have the following configuration:

---------------

total nodes => 16 

gpu_count => 0

cores_per_node => 32

memory size => 754.37 GB

product name => Dell PowerEdge R640

SLURM CPU constraint => --constraint=CPU-Gold-6130

SLURM mem setting => --mem=754000



New Software Available

Thanks to the addition of the new GPU nodes, we can now support software many of you have been requesting.  TensorFlow, pytorch, and keras are available now and more software capable of utilizing the new GPUs will be installed over the coming months.  To see what software versions are available, use the command:  module avail



OnDemand

OnDemand has been upgraded to version 1.3.  Users have a few more options when submitting jobs now, such as being able to specify how many nodes they want their job to launch on.  We’ve had many requests to run Jupyter Notebook on the cluster so we’ve added a separate app to do this easily.  Industry users also have access to a remote desktop app that runs on the industry nodes.  More features will be introduced in the coming months. 



Coldfront

Coldfront version 2 will soon be released. The product has been completely refactored from the ground up and offers a more streamlined process for creating accounts. We are also happy to announce we will be sharing this with the open source community this fall. More details will be provided in the next few weeks. While we transition, we ask faculty to send any requests for account additions or removals via email to CCR help or submit a ticket in the help portal. Any emails you received about renewing your subscriptions can be ignored until the new Coldfront is online. Nothing will happen to your CCR accounts or access during this time period.  As usual, we ask faculty who are teaching classes to wait until the drop/add period is over before requesting CCR accounts.