Blog

Adding support for AI workloads accelerated by NVIDIA infrastructure

More and more users have requested to control the default Docker engine configuration so that they can take advantage of NVIDIA GPUs installed on their bare metal servers. We are excited to announce that we have extended the platform to support AI workloads running GPU-accelerated Docker containers. Let’s dive into this new feature to go through on: What is it? Why did we build it? How can you use it?

If you are not familiar with GPU-accelerated containers, check out NVIDIA’s Overview of cloud native technologies documentation to learn more.

New unit: Docker runtime NVIDIA

From the start, Cycleops is a deployment tool built on the concept of automation units, which allows our customers to design their desired software stack and within a few minutes deploy their Dockerized applications without worrying about writing scripts. We added the “Docker runtime NVIDIA” unit to automate the tasks of installing the required container toolkit and configuring the Docker runtime so that our customers can run their AI applications with GPU-accelerated support. That new unit is purpose-built to work well with the set of Docker engine and Docker container units provided with Cycleops, nevertheless can be applied to any pre-installed Docker setup that is known to be operational.

AI workloads are quite popular

One of the major benefits of working with Docker containers is that the application can run without dependencies from the host operating system that makes portability much easier. With the rise and increased popularity of AI technologies during the last years, developers have introduced a new requirement for running pertinent applications using hardware acceleration. At the same time Docker is the de facto standard for modern software development and amazingly supported by GPU vendors as NVIDIA. Our team is always on duty to adhere to new industry requirements and fulfill our mission to make deployments simple for all teams of all sizes!

Give it a run with Cycleops

If you’d like to learn more about simplifying Docker deployments with the Cycleops no-code automation tool, check out our documentation of designing your stacks and deploying your setups.

💡Interested in trying Cycleops? Start for free today!

Exit mobile version