Kubernetes v1.31: Optimized CPU Distribution for Enhanced Performance on Multi-Core Processors

In Kubernetes v1.31, we are excited to introduce a significant improvement in CPU management: the distribute-cpus-across-cores option for the CPUManager static policy. This feature is currently in the Alpha phase and is hidden by default. It marks a strategic shift aimed at optimizing CPU utilization and enhancing system performance on multi-core processors.

Understanding the Feature

Traditionally, Kubernetes’ CPUManager tends to allocate CPUs as compactly as possible, typically packing them onto the smallest number of physical cores. This allocation strategy is critical because CPUs on the same physical host continue to share some resources of the physical core, such as cache and execution units.

While the standard approach minimizes inter-core communication and can be advantageous under certain scenarios, it also presents a challenge. CPUs sharing a physical core can lead to resource conflicts, resulting in performance bottlenecks, particularly noticeable in CPU-intensive applications.

The new distribute-cpus-across-cores feature addresses this issue by changing the allocation strategy. When this option is enabled, the CPUManager policy instructs the distribution of CPUs (hardware threads) across multiple physical cores as much as possible. This distribution aims to minimize conflicts between CPUs sharing the same physical core, thereby improving application performance by providing them with dedicated core resources.

Technically, within this static policy, the list of available CPUs is reordered as depicted in the diagram to allocate CPUs from separate physical cores.

Enabling the Feature

To enable this feature, users must first add the Kubelet flag --cpu-manager-policy=static or the field cpuManagerPolicy: static in the Kubelet configuration. Then, the option --cpu-manager-policy-options distribute-cpus-across-cores=true or distribute-cpus-across-cores=true can be added to their CPUManager policy settings in the Kubernetes configuration. This setting instructs the CPUManager to adopt the new distribution strategy. It is important to note that this policy option currently cannot be used in conjunction with the full-pcpus-only or distribute-cpus-across-numa options.

Current Limitations and Future Directions

As with any new feature, especially one in the Alpha phase, there are limitations and areas for future improvement. A significant current limitation is that distribute-cpus-across-cores cannot be combined with other policy options that may conflict with CPU allocation strategies. This limitation may affect compatibility with certain workloads and deployment scenarios that rely on specialized resource management.

However, we are committed to improving the compatibility and functionality of the distribute-cpus-across-cores option. Future updates will focus on resolving these compatibility issues, allowing this policy to seamlessly integrate with other CPUManager policies. Our goal is to provide a more flexible and robust framework for CPU allocation that can adapt to a variety of workloads and performance requirements.

A practical example of utilizing this new feature could be a company running compute-intensive applications, such as data analytics or machine learning. By enabling the new distribution policy, these applications can benefit from improved CPU performance, leading to faster results and more efficient operations. At ayedo, we support companies in optimizing their use of Kubernetes and integrating these new functionalities into their workflows.

Source: Kubernetes Blog