Get Ready for Delta

NCSA continues its contributions to supercomputing innovation with the coming launch of Delta, a computing and data resource that balances cutting-edge graphics processor (GPU) and CPU architectures that are shaping the future of advanced research computing. With 848 GPUs, Delta will be the most performant GPU resource in the National Science Foundation portfolio when it officially launches.

The hardware was installed in December 2021 and is currently being readied for early user testing, says Brett Bode, assistant director for the Blue Waters Project Office and a co-PI on the Delta Project. Allocation requests are being accepted and will begin after the machine deploys.

“Delta is one of the first non-Cray Shasta systems to use the HPE/Cray Slingshot interconnect. As with many new technologies, there have been some issues that had to be addressed during system installation that extended the planned installation time. Those have been addressed and NCSA staff are now hard at work installing the user space software and testing the system in preparation for the early user period,” explains Bode.

HPE Slingshot is a high-performance network for HPE Cray supercomputers and HPE HPC clusters designed for running diverse simulation, modeling, AI, and analytics workloads on one system. According to the HPE Slingshot website, cloud, AI, and IP-based applications and frameworks are able to run without modification, “enabling a new class of users to transition from traditional cloud-based computing to take advantage of supercomputing performance, capabilities, and scale.”

Is Delta right for you?

Delta might be right for you if:

• You need access to the latest NVIDIA GPUs for your accelerated code.
• You’re interested in transitioning from CPU to GPU while running a CPU workload, for example if you want to scale up your models.
• You have large memory demands for a shared-memory application such as in-memory databases, etc.
• You have a very I/O-intensive application that would benefit from an exceptionally fast flash-based file system.

Easy to use

When deployed, Delta will boast a non-POSIX file system that presents a POSIX-like interface, allowing applications to reap the benefits of modern file systems without rewriting code. 

“A recent study by the University of Illinois Urbana-Champaign and Lawrence Livermore National Laboratory showed that many HPC applications do not require strong POSIX consistency semantics. Delta’s relaxed POSIX semantics file system provides immediate I/O performance gains for those applications,” says Greg Bauer, senior technical program manager for NCSA’s Scientific and Engineering Applications Support group and a co-PI on the Delta Project. 

Research computing has grown from GPUs being used as accelerated computing support for special applications to GPUs being used for a broad mix of scenarios, including hybrid computing, machine learning, accelerated computing, visualization and user-interactive computing.

It’s Time to Prepare

Will you be ready to take full advantage of Delta when the system comes online?

This Introduction to Delta webinar shares details about Delta and how to get research codes up and running. The webinar is led by Tim Boerner, senior assistant director at NCSA and lead for the Delta Project Office, Bode and Bauer, all co-PIs on the Delta Project.

NVIDIA training materials are also available. 

Delta Allocation Award Opportunities

NCSA will integrate Delta into the national cyberinfrastructure ecosystem through the Extreme Science and Engineering Discovery Environment (XSEDE) and partner with the Science Gateways Community Institute to provide platform access serving a broad range of needs. 

The majority of time on Delta will be allocated by XSEDE, which will award 90-percent of Delta’s cycles to the national community of researchers and educators. 

Five percent of Delta is reserved for researchers at the University of Illinois Urbana-Champaign and is awarded two times per year through a competitive, campus allocation process. There is an additional five percent of the system reserved for innovative/exploratory, diversity, Science Gateways and data science purposes.

A portion of NCSA’s advanced computation systems have traditionally been dedicated for researchers at Illinois. It was a popular feature of the Blue Waters Project that we are now portion of NCSA’s advanced computation systems have traditionally been dedicated for researchers at Illinois. It was a popular feature of the Blue Waters Project that we are now excited to be able to offer through the Delta Project as well. Researchers can request an XSEDE or Delta Illinois allocation by following procedures found on our allocations web page. We’re excited to offer the advanced research capabilities of Delta to the Illinois and national academic research community.

Laura Herriott, associate director for User Services at NCSA