Luna provisions just-in-time right-sized cost-effective compute for your app when your app starts and terminates the compute when the app terminates.
ML workloads like Ray are bursty and use GPU devices which are expensive to waste - this makes them a great fit for Luna. Check out our CNCF blogpost to see how Luna powering a GPU Ray cluster with a GPU-enabled Ray Head reduced elapsed time by 61%, computing cost by 54%, and idle Ray cluster cost by 66%, while retaining the performance quality of the AutoML results and reducing operational complexity.
Build/test workloads are bursty in nature and require predictable resources. Luna is a great fit for CI since it eliminates wait times during unexpected spikes and prevents wasted spend on unused capacity during downtimes (ex: nights, weekends). Luna works seamlessly with any CI orchestrator (Buildkite, CircleCI, GitHub Actions, Jenkins, etc).
The availability of mac1.metal instances on AWS is a game-changer in consolidating Mac workloads on the cloud. Luna eliminates DevOps overhead associated with: