Supercomputing Resources | VCR | University of Nebraska Medical Center

Following are High Performance Computing Resources offered to the UNMC Research community.

GPU based A100-DGX system
CPU based INBRE cluster
Holland Computing Center (HCC) at Peter Kiewit Institute
National Super Computing Network (XSEDE)

GPU based A100-DGX system

This super computer is part of Center for Intelligent Health Care (CIHC) under the leadership of Dr. John Windle. The RITO has deployed the system at UNMC, provides the service, and support maintenance of the system.

The system is very powerful resource for running Artificial Intelligent (AI), Machine Learning (ML) Algorithms, and Deep Learning (DL) neural networks that can handle analysis of large datasets. In addition, this system is one of a kind in NE and it is hosted in UNMC HIPAA compliant environment. Researcher can make a request to have an account on the system and perform computing jobs.

System specifications:

One node:

CPUs: 2x64x2 sockets/cores-per-socket/threads-per-core, AMD EPYC 7742 64-Core Processor
GPUs: 8, nVIDIA A100 Tensor Core, 40 GB GPU memory each
RAM: 1008 GB
Local Storage: 15.58 TB solid state

Ubuntu Linux 20.04 with standard development packages and libraries installed. SLURM job scheduling and Docker container support are available.

Generally the researcher is responsible for installing their applications in their own home directory, however, we can install general purpose apps to be used and shared system-wide that would benefit the user community. Examples of that would be Java Runtimes, Python, MATLAB, R, SAS, TensorFlow, etc.

Resources:

NVIDIA Catalog - https://ngc.nvidia.com/catalog
Hands on Training - https://www.nvidia.com/en-us/training/
GPU Accelerated Apps - https://www.nvidia.com/en-us/gpu-accelerated-applications/

CPU based INBRE cluster

Nebraska INBRE sponsored two cluster with attached storage has been deployed by the RITO in UNMC HIPAA compliant environment. The RITO provides the service, and support maintenance of the system. The UNMC users and INBRE associated Primary Undergraduate Institutes researchers can make a request to have user accounts on the system by emailing to rito@unmc.edu

System Specifications (224 cores / 448T):

Two nodes each with 112 cores/224 Threads. Both nodes are mounted onto dedicated storage servers.

Each Node – DELL POWEREDGE FC630 with dual Intel Xeon 2.3 GHz CPU, 14C/28T, 256 GB RDIMM memory, 10 Gbit/sec network connectivity.

One storage server can host up to 20 TB data and second storage server can host upto 36 TB of data. Both these servers are connected to INBRE computing nodes on 10 Gbit/sec network fiber.

Holland Computing Center (HCC) at Peter Kiewit Institute (PKI) in Omaha

The Research IT Office (RITO) has close working relationship with PKI. If you are planning on accomplishing some work on the supercomputers at PKI, please contact us. The UNMC user can request a user account on the HCC cluster to perform High-Performance Computing data analysis using non-PHI data. HCC users can access and use many HCC resources in a shared manner at no charge:

Crane and Rhino are HCC’s high-performance clusters for general usage. A user can run jobs on these clusters without charge. As these are shared resources, there are some limitations on jobs. For example, by default, the max job run time is 7 days, the max CPUs available per user is 2000, and the max job number per user is 1000. Details on these limitations can be found here. The general cluster documentation can be found here.
Anvil is HCC's cloud platform. Each research group can initially spawn 10 virtual machine instances with a combined total of 20 virtual cores, 60GB RAM, 10 storage volumes, and 100GB volume storage capacity. The detail of per group limits can be found here.
For storage, users can access three main filesystems including Home with a 20GB quota per user, Common with a 30TB free quota per research group, and Work with a 50TB quota per group. Additional details on the limits and usage expectation of these filesystems can be found here.

Again, as these are shared resources, certain limitations are applied. HCC also offers Priority Access on a cost-recovery basis to the HCC user community. Details on Priority Access can found at here.

XSEDE (Extreme Science and Engineering Discovery Environment)

It is a cyber infrastructure distributed across the US. XSEDE is a single virtual system that scientists can use to interactively share computing resources, data, and expertise. The XSEDE infrastructure has been established using funds from NSF ($121 Million). Using XSEDE, researchers can establish private, secure environments that have all the resources, services, and collaboration support they need to be productive. Visit here to see all resources offered by XSEDE. XSEDE supports 16 supercomputers and high-end visualization and data analysis resources across the country. It also includes other specialized digital resources and services to complement these computers.

UNMC has an XSEDE membership that makes it easy for all UNMC researchers to get free access to XSEDE supercomputers. If you want to apply for the user account, follow the instructions listed in this document.

******* Free On Demand Online Training for Supercomputer Users ********

High quality online workshops are available from Cornell University to all XSEDE supercomputer users. Please setup your free user account on XSEDE by following attached document. After creating user account (it will take about 1 business day), please go ahead and jump into the workshop material.

Here is the list of courses (current as of November 2013) from the workshop home page. Visit the workshop home page for the most up-to-date list of course offerings.

Comprehensive Courses