Posted On: Mar 7, 2022

Bottlerocket, a Linux based operating system that is purpose built to run container workloads, now supports GPU based EC2 instance types that are powered by NVIDIA. Customers can now benefit from using the same container focused host operating system for both their non-GPU and GPU workloads, including machine learning, video encoding, and streaming workloads. This helps customers standardize on a single operating system that utilizes the underlying specialized compute hardware.

Lack of support in Bottlerocket for specialized processors required customers to use a different operating system for GPU based instance types powered by NVIDIA. Customers needed to build and deploy custom AMIs to install additional software, and make additional configurations to use the underlying GPU resources. This was operationally intensive, time-consuming, and required customers to maintain multiple workflows. Some customers choose to perform these host configurations at runtime. However, this increased node launch times when using auto-scaling or EC2 spot instances and introduced additional dependencies on external tools and repositories.

With Bottlerocket adding support for GPU based EC2 instance types powered by NVIDIA, customers can reduce operational and cognitive overhead in managing container hosts with different operating systems bringing consistency of operating system experience. The new Bottlerocket AMI includes the necessary software components required to run containerized accelerated workloads built into the base image. This enables secure, seamless installation of required NVIDIA drivers and its updates, improves time to node ready state, and reduces dependencies on external tools and repositories.

Customers can choose the new Bottlerocket AMI using EKS self-managed nodes and Karpenter on EC2 instances powered by NVIDIA. The AMI is available in all AWS commercial and GovCloud regions at no additional cost. You only pay for the EC2 instances that you use. Please refer to our blog to get started. Learn more about Bottlerocket at our product page and Bottlerocket’s Github repository.