In cloud-native environments, application performance often degrades due to contention over shared resources such as CPU caches and memory bandwidth. Current container technologies lack mechanisms to isolate these resources, which compels operators to maintain low utilization by scaling out their deployments.

This session explores strategies used by hyperscalers like Google and Alibaba Cloud to mitigate such performance interference. We will review their published methodologies, extracting key principles that could guide the development of a Kubernetes-native performance isolator. Participants will gain insights into the design trade-offs and operational impacts of these tools. Additionally, we will discuss integration strategies for deploying such isolators in existing Kubernetes environments, aiming to optimize resource utilization while preserving application performance.