BalancedResourceAllocation scheduling priority¶
BalancedResourceAllocation is a default scheduling priority strategy in Kubernetes used during the Priorities phase of the scheduling process^[400-devops__06-Kubernetes__k8s-paas__原理及源码解析__Kubernetes调度机制.md]. Its primary goal is to ensure that resources are allocated evenly across different types (e.g., CPU and memory) on a node, rather than allocating all of one resource type while leaving other resource types idle.
Mechanism¶
In the scheduling workflow, this priority function works alongside the [[leastrequestedpriority|LeastRequestedPriority]] strategy^[400-devops__06-Kubernetes__k8s-paas__原理及源码解析__Kubernetes调度机制.md]. While LeastRequestedPriority favors nodes with the most total free resources, BalancedResourceAllocation focuses on the balance between those resources^[400-devops__06-Kubernetes__k8s-paas__原理及源码解析__Kubernetes调度机制.md].
The score is calculated using the following formula:
score = 10 - variance(cpuFraction, memoryFraction, volumeFraction) * 10
Where "Fraction" represents the ratio of the Pod's requested resources to the node's available resources^[400-devops__06-Kubernetes__k8s-paas__原理及源码解析__Kubernetes调度机制.md]. The scheduler uses a variance algorithm to measure the "distance" between these resource fractions. The node with the smallest gap between resource fractions receives the highest score^[400-devops__06-Kubernetes__k8s-paas__原理及源码解析__Kubernetes调度机制.md].
Objective¶
The main objective of this strategy is to prevent resource skewing, such as a scenario where a node's CPU is fully allocated while its Memory remains mostly unused^[400-devops__06-Kubernetes__k8s-paas__原理及源码解析__Kubernetes调度机制.md]. By selecting the node with the most balanced resource allocation, the scheduler optimizes resource utilization across the cluster.
Related Concepts¶
- Kubernetes Scheduler
- [[LeastRequestedPriority]]
- [[QoS]]