Who is Sylabs? We are often asked about who Sylabs is, and what we are working on for both the open source community, and for those who require professional support of their HPC container workflow environments. Sylabs was created in 2017 to provide support,...
TL;DR: Enterprise adoption of RHEL 8 is a transition over time, not a discrete event in time. Singularity containers preserve your heavily vested legacy deployments, while enabling you to make the transition on your terms.
Yesterday, general availability of Red Hat Enterprise Linux 8 was announced at the Red Hat Summit in Boston. As you’d expect for a major version increment, this is a significant release. For many enterprise customers, there is always a degree of ambivalence associated with such advancements: On the one hand, they are understandably keen to benefit from all that the new version has to offer; yet at the same time, this enthusiasm must consider the conservative needs of the enterprise first.
Well beyond matters of just managing change responsibly in their production enterprise deployments, many enterprise customers require the entire toolchain to be certified on a per-platform basis. This is particularly true for organizations engaged in engineering, in which platform-specific certification is critical to their ability to obtain reliable results from the design-automated modeling and simulation that they perform in their data centers or in the cloud – computer-mediated automation that has replaced physical simulations in the laboratory in many cases.
Consider the following representative example. Physically crashing automobiles into obstacles at different speeds, angles of incidence, environmental conditions, etc., used to be a routine component of the engineering process. With the advent of mechanical-design automation however, the requirement for these ‘physical experiments’ has been greatly reduced, as matters such as vehicle safety and quality can be assessed through parameter studies based upon modeling and simulation performed in the data centers of these manufacturers. There’s just one catch: results obtained are only certified as valid for specified platforms – literally the toolchain stack detailed from the application and its workflow, down to the operating system and hardware, and everything in between … in some cases, with surprising degrees of granularity!
In other words, for these organizations, introducing a brand new operating system such as RHEL 8 is much more of a transition over time, as opposed to an event in time. Reliable results in support of the engineering process aren’t a nice to have, they’re a fundamental business requirement for any company that places premiums on product safety and quality – and ultimately the competitiveness and viability of their business itself over the long term.
Containers provide the means to package up an application along with all of its dependencies and environment – the binaries and libraries in its toolchain, configurations, data – everything except the operating system kernel. By removing the explicit dependency on a particular implementation of the Linux kernel, containers provide the means to ensure that the adoption of RHEL 8 can proceed effectively and efficiently as a transition over time, as opposed to an event in time.
In fact, the value proposition of containers is so compelling in this regard that it is often support for ‘some legacy configuration’ that is responsible for their introduction in the first place:
When we first adopted Singularity a few years ago, the driving case was to support TensorFlow on our cluster with GPUs and an old operating system. It was relatively easy to build a container with the right libraries and environment using Singularity.
Chris Reidy, principal HPC systems administrator Research Technologies at the University of Arizona
In other words, the inevitable change inherent in the underlying operating environment did not negatively affect a previously existing investment at the U of A in a known-good configuration – owing to use of Singularity containers.
RHEL 8 Linux Containers
Secure containers for the needs of the enterprise; that’s the positioning you’ll find in the RHEL 8 announcement. Along with their offering that targets microservices based applications and workflows, we wanted to remind you that Singularity containers offer a general-purpose alternative for microservices as well as compute-driven workloads. Also daemonless in terms of implementation, Singularity ups the security ante significantly by allowing for cryptographic signing and verification of certified images for enterprise-grade use cases – so engineers know they’re making use of the correct container image in their crash simulations, for example.
In addition to security, Singularity containers deliver close to bare-metal characteristics after startup, so the solution is also performant in compute-intensive engineering use cases. From the availability of ancillary services in the cloud to standards compliance, interoperability with other container platforms (e.g., Docker), and finally integration with Kubernetes, Singularity containers offer industry leading flexibility without compromising security.
Given that the founder of Singularity was previously responsible for founding 15-year old CentOS Linux, you can rest assured that your systematic transition to RHEL 8 is something the team at Sylabs is uniquely qualified to enable. In fact Gregory Kurtzer, and a team from Sylabs, can be found at booth #1133 today at the Red Hat Summit in Boston. Whether it’s transitioning to RHEL 8 with your needs in mind, or adopting containerization owing to some other driver, we encourage you to take advantage of the opportunities available to you in Boston today at the Red Hat Summit.
The user, developer and provider community around the open source implementation of Singularity make routine use of RHEL, CentOS and other Linux distributions. Our testing matrix for SingularityPRO includes ‘legacy’ RHEL, as well as CentOS and SLES. While we together gradually embrace RHEL 8, there’s never been a better time to factor Singularity containers into your PoC to production deployments within your organization, and we look forward to engaging with you in that conversation.
Join Our Mailing List
Now that we have readied the WSL2 environment with Singularity and the relevant CUDA libraries, it’s time to run the sample Keras workflow.
Signing the Container The Singularity 3.0 family introduced the ability to create (and manage) PGP keys to sign and verify containers. This provides a trusted method for Singularity users to share containers and ensures a bit-for-bit reproduction of the original...