Spotinst Ocean ensures pods and tasks are placed on the best possible VM that fits their resource requirements, along with other scheduling constraints such as labels, attributes, taints, tolerations, and affinity rules.
However, estimating the proper amount of CPU and Memory when assigning resource requests to deployments or services is also a challenge which teams are facing when designing Kubernetes or ECS clusters.
To address this challenge and create even more resource-efficient clusters, Spotinst had implemented a Right Sizing suggestions mechanism, as an integrated part of the Ocean solution.
Spotinst Ocean, empowered by Right Sizing, monitors Kubernetes Pods or ECS tasks in real-time and provides recommendations for adjusting the resource requirements each workload type (deployments for Kubernetes, or Services for ECS).
With Right Sizing, Ocean continuously monitors the cluster’s CPU and Memory usage and provides notifications when a workload consumes significantly more/less resources than it requested.
Applying the changes suggested by those notifications will help utilize resources in the cluster in a more precise manner, and lower chances of cluster issues as resulting from under\over utilization of resources.
- For Kubernetes Ocean clusters, Right Sizing relies on the Kubernetes Metrics Server. For more details see the following tutorial.
- For ECS Ocean clusters, Right Sizing relies on Cloudwatch metrics, and displays data after an initial 4 day data collection period.
Right Sizing Suggestions can be viewed on your Ocean Dashboard, under the Right Sizing tab, and in the detailed pages for Kubernetes Deployments or ECS Services.