Eduardo packs a lot into this presentation. However, the overall message is straightforward: Singularity is simply a better fit for Enterprise Performance Computing (EPC) than any other container option available. Owing to his passion for the technical merits of containerization, Eduardo objectively makes the case for Singularity.
You’ll want to review the entire presentation to extract all of the details. To whet your appetite, consider the following:
- Singularity handles 40 GB images (and even much larger) – as some use cases call for very large containers
- Singularity emphasizes integration over isolation – as this makes ‘the hard stuff,’ like GPU and/or InfiniBand/OmniPath access, a breeze
- Singularity always emphasizes security – while allowing untrusted users to run untrusted containers securely
- Singularity encapsulates the container runtime into a single file – Singularity Image Format (SIF) is elegantly simple yet highly mobile and secure, as images are signable and verifiable
As noted in a recent post (please refer to the section regarding “Nomad Support”):
Nomad is already factoring into some leading-edge use case examples.
In his presentation, Eduardo provides an image-processing demonstration for edge detection involving a plugin he has been developing. This plugin integrates the Nomad runtime with Singularity directly – thus enabling Singularity as a container runtime under Nomad. Because Eduardo is currently putting the wraps on this plugin, please stay tuned for more on direct use of Singularity containers via Nomad.
We’ve embedded Eduardo’s HashiConf presentation below for your convenience. Should you wish to review the abstract for his presentation, or gain easy access to others given at HashiConf 2018, please click here.