skip to Main Content


TL;DR: Second SUSE booth presentation from SC18; video here. NVIDIA supernova demo CLI deconstruction here. Singularity containers for compute-driven workloads in the Enterprise on SLES.

Even though the dust has finally settled from his supernova demonstration at SC18, NVIDIA CEO and founder Jensen Huang’s captivating moment remains exceedingly memorable. In this second presentation recorded on the exhibits floor at SUSE’s booth, Sylabs’ Ian Lumb deconstructs Huang’s supernova demo from a technical perspective. Getting down to the command-line syntax (see Slide 7 here), Ian explains how the entire demo was literally contained by Singularity, in a 7 TB, real time ParaView visualization that spanned multiple GPUs.

Ian leveraged the NVIDIA supernova demo to illustrate a compelling example of Enterprise Performance Computing (EPC) – a relatively new class of workload that places emphasis on computation over, for example, the delivery of services. Whereas Huang’s example focused on astrophysics, there are numerous examples of real-time visualization that are more directly Enterprise in nature – for example, visualizing in 3D an engineered prototype from a bicycle to an airplane.

In addition to deconstructing the NVIDIA supernova demo to illustrate EPC, Ian covers the following topics in his presentation:

  • The not at all subtle differences between Singularity and other implementations of containers
  • The compelling, Enterprise-grade combination of SUSE Linux Enterprise Server (SLES) and Singularity containers
  • The Sylabs Cloud offering – a Singularity-centric portal hosted in the cloud that enables key signing and verification, hosted containers, and remote builds
  • Use cases that blend traditional HPC with services-oriented workloads in a demanding hybrid fashion
  • The short-term roadmap for Singularity containers – a roadmap that features standards compliance plus integrations with Kubernetes and HashiCorp’s Nomad

Both of Ian’s presentations illustrate the ease with which GPUs can be exploited when executing Singularity containers – i.e., by little more than including the “–nv <container URI>” option when executing a Singularity “run” command. Such is the expected outcome of emphasizing integration over isolation when it comes to containerization.

You can watch Ian’s presentation below or via the SUSE Youtube channel here. The slides he used are available here.

Kudos and thanks again to our great friends at SUSE for recording booth presentations, making the industry’s best parody videos, and just being a whole lot of fun to work with. We look forward to SC19 in Denver and working with them again in about 300 days!

Back To Top