NVIDIA CEO Jensen Huang is known for showcasing spectacular demos in any presentation he delivers. At SC18, he did just that - again! This time, the NVIDIA team provided an interactive visualization of a supernova event in the dwarf galaxy M82. Because this is an interactive visualization, ParaView is required to manipulate a 7 TB dataset in real time. Whereas the demo is visually stunning (of course!) and scientifically advancing, we wanted to let you know about it because the entire demo made use of Singularity containers. Culminating ultimately in a “singularity run” of a Docker container for ParaView in the NVIDIA GPU Cloud (NGC), this demo was executed across multiple GPUs via MPI. In Huang’s own words:
That is absolutely the best container demo you have ever seen!
We couldn’t agree more - it truly was “... container marketing at its best!”
Singularity containers for ParaView are now routinely available via the NVIDIA GPU Cloud. To quote Huang again:
This is not the app store of apps you want; this is the app store for apps you need.
And in this ‘app store’, it’s exceedingly clear from the supernova demo, as well as Huang’s invited presentation, that Singularity is the obvious choice for containerizing the most demanding compute driven workloads that matter to the HPC community. Many have argued that enterprise customers follow the lead of SC18’s HPC elite - and that this has been going on for 30 years now! HPC inspires! No wonder then, NVIDIA is seeding the ‘new HPC’ with proven, mobile, and highly performant application stacks via Singularity containers:
Nice and simple. Download it, grip it, and rip it!
(Huang really has a way with words!)
Huang’s marathon of a presentation runs almost two hours! We highly recommend you watch every minute. The supernova demo containerized via Singularity is where we’ve redirected you in the video below. We are delighted to have NVIDIA’s endorsement of the entire container ecosystem, and Singularity in particular - a point similarly endorsed by the audience’s applause during this part of Huang’s presentation. Possibly the best part of this entire post is that, really for the first time, you can reproduce demos like this one for yourself - using the same Singularity containers provided by NVIDIA, on data that matters to you. All you need are a few GPUs and Singularity to get started. We look forward to seeing what you produce.