I would like to....

DATA COMPUTE

Alpine

Alpine is a heterogeneous supercomputing cluster based primarily on the AMD EPYC “Milan” CPUs with; HDR InfiniBand; and 25 Gb Ethernet. Alpine is jointly funded by the University of Colorado Boulder, the University of Colorado Anschutz, Colorado State University, and the National Science Foundation (award 2201538).

Get Started with alpine

Transition from Summit to Alpine

Alpine is now open to all CSU users. If you are a current Summit user, you do not need to request an account for Alpine, you automatically have access.

Summit

System Specs: With a peak performance of over 400 TFLOPS, there are 472 general compute nodes, 10 GPU nodes, 5 high-memory nodes, and 20 Intel “Knights Landing” nodes.  All nodes are connected through a high-performance network based on Intel Omni-Path with a bandwidth of 100 Gb/s and a latency of 0.4 microseconds. A 1.2 PB high-performance IBM GPFS file system is provided for active computation data.

For more information, see the Summit User Guide or learn more About Summit. 

DATA TRANSFER 

Data transfer needs take many forms. Here we address common data set transfers to and from the Alpine system. See CU Boulder’s Data Transfer page for other file transfer methods supported on Alpine and for a detailed step-by-step process to initiate the file transfers on Alpine. Please let us know at [email protected] what other forms of data transfer you would like to see added here.

SFTP

The sftp utility is an interactive alternative to scp that allows multiple, bi-directional transfer operations in a single session. Within a sftp session, a series of domain-specific file system commands can be used to navigate, move, remove, and copy data between a local system and Research Computing resources.

Good for small data sets (<25MB) i.e batch script, small files, error logs, etc.

Good for medium data sets (25MB to 10GB) i.e codes, software files, libraries, small datasets, output files, etc.

Globus

Globus transfer can be used on MacOS, Windows, and Linux OS and is research computing’s recommended way of transferring data. The file transfer is initiated through an interactive web application. See the Get Started with Globus page to learn more.

Good for large data sets (over 10GB) i.e. large output files, large datasets, large models, etc.

DATA STORAGE

RSTOR

RStor is a centrally provided, scalable network-attached storage platform by IX Systems and it currently has a capacity of 879 Terabytes. For data protection, the RStor system creates daily read-only point-in-time copies (snapshots) of files for 30 days.  In addition, a nightly disaster recovery (DR) backup is taken of the entire system to different locations on campus. Each user is encouraged to develop and implement a data archiving practice.  

The Office of the Vice President of Research in partnership with the Division of IT has implemented a storage solution for researchers needing a secure location for their working data. Currently configured with 879 terabytes of storage, this IXSystems’ TrueNas storage is located in the campus data center and can be accessed directly from Windows or Unix-based systems. As usage has grown, the system has been expanded to meet the capacity demands of the campus.

RStor solves the problem of access to high-volume, scalable storage of unstructured data for research needs, and through a hybrid central funding model it is cost-effective for storing large volumes of research data. An additional 50 drives can be added quickly to almost double the capacity of the system to 1.6 Petabytes with research demand. This can further be expanded strategically to 3.265 Petabytes utilizing the storage server infrastructure and adding additional enclosures to maximum usable disk capacity.

The Office of Vice President of Research is subsidizing this system, allowing for this iteration of the system to be available to researchers at $80 per terabyte per year.   Space can be requested through your local college IT department, which will manage billing, quotas, and any drive mappings you require. 

If your college or department does not have a pre-registered administrator, and you have researchers wanting to use the system, please contact [email protected].

Please visit rstor.research.colostate.edu to find your local IT department’s point of contact for this service.

PetaLibrary

The PetaLibrary is a University of Colorado Boulder Research Computing service that supports the storage of data interconnected with the Summit HPC system.