Dell Technologies introduced open source Omnia software this week to spotlight its latest initiatives for easing the deployment and management of AI, analytics and high-performance computing applications.
The Dell Technologies HPC & AI Innovation Lab collaborated with Intel and the HPC community on the Omnia software with the goal of helping users automate the provisioning and management of AI and compute-intensive workloads by creating a single pool of flexible resources to address changing needs.
Caitlin Gordon, vice president of product management for integrated solutions at Dell Technologies, said Omnia would enable users to employ the same infrastructure for different types of jobs and put the "right software" on each server based on the use case, such as HPC simulations or neural networks for AI.
"You'll be able to reduce your time of deployment from weeks down to hours," Gordon said. "So, it's all about flexibility. One of the coolest things about Omnia is that you can compose and recompose the stack based on what you need, really taking an infrastructure-as-code approach here."
The Omnia software stack combines a set of deployment-speeding Ansible playbooks, the Kubernetes container orchestration engine, Slurm workload manager, library frameworks, services and applications. Gordon said the Omnia software has been available for a few months on GitHub. Contributors to the open source Omnia project include Arizona State University Research Computing.
Dell HPC, AI and analytics focus
Addison Snell, CEO of Intersect360 Research, said Dell's Omnia release falls in line with market trends in diversification, including workloads mixing AI, data analytics, and traditional scientific HPC and technology elements such as specialized processors and accelerators.
"Omnia is designed to be a common, consolidated environment to ease deployment and management of clusters across these expanding needs. By making it open source, they allow for customization, like a sandwich shop that allows a customer to amend their selection," Snell said. "In this way, Omnia also follows the trend we see toward composability, allowing for the selection of different modules to suit the tasks at hand, including through the use of containers."
Snell said he expects Omnia to appeal most to organizations looking for supported environments that can span a wide range of high-performance workloads. He said Intersect360 surveys show that 80% of HPC users have either incorporated machine learning into their environments or are actively working to do so, usually on shared or common infrastructure. Dell designed Omnia to target those types of environments, Snell said.
Steve McDowell, a senior technology analyst at Moor Insights & Strategy, said it was no surprise to see Dell release the Omnia software to help ease deployments on its servers and storage because the company has increasingly focused on HPC and AI over the past several years.
"HPC, supercomputing and AI are all workloads that treat infrastructure as fungible, meaning the underlying hardware is repurposed and reconfigured to meet the needs of the application," McDowell said. "The industry has largely adopted Ansible as the mechanism to dynamically configure and reconfigure infrastructure. It's natural that Dell is delivering these capabilities as Ansible playbooks."
McDowell said Dell is not alone in taking the Ansible approach. He noted that IBM and HPE, for example, support the HPC and AI community with Ansible playbooks.
Dell HPC on demand extends to VMware
Also in connection with this week's ISC High Performance 2021 Digital event, Dell extended its HPC on-demand offering to VMware's Cloud Foundation, Cloud Director and vRealize Operations. HPC on demand is designed to give customers fast access to resources and let them pay only for what they use.
Gordon said the addition of VMware support to Dell HPC on demand would give Dell the ability to not only offer HPC on demand in a single-tenant environment but also in multi-tenant environments.
"It's really about being able to support our customers who may not have the resources or may not want to invest in their infrastructure and people because their HPC workloads maybe are inconsistent. Or, maybe they need to complement what they already have in their environment with a secondary environment to burst to," Gordon said.
McDowell said that traditional HPC management tools often are not the same tools that enterprises use to manage their IT infrastructure. So, the addition of VMware as a management option for Dell's supercomputing on-demand offering should make it much easier for traditional enterprise IT administrators who are already comfortable using VMware to easily manage HPC workloads, he said.
Also this week, Dell announced that it now offers options for Nvidia A30 and A10 Tensor Core GPUs with Dell EMC PowerEdge R750, R750xa and R7525 servers to address AI and deep learning workloads. Gordon said ideal use cases include VDI, inferencing and interactive rendering.
Carol Sliwa is a TechTarget senior writer covering storage arrays and drives, flash and memory technologies, and enterprise architecture.