Kubernetes Resource Scheduling via Simulation-Based Reservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing resource scheduling methods in Kubernetes often fail due to nodes that cannot meet the resource requirements of pods, leading to a low success rate of resource scheduling.
Innovation Solution
A resource processing method that creates instances corresponding to the number of pods needed, performs simulation scheduling across multiple nodes to determine target nodes with reserved resources, and generates a resource reservation result to instruct resource scheduling using the reserved resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If default scheduling method is used to deploy pods to nodes, then scheduling process is simple, but resource scheduling success rate is low
Solution Approach 1:
The patent performs simulation scheduling before actual resource scheduling to pre-determine target nodes and reserve resources. This preliminary action ensures that when actual scheduling occurs, the resources are already confirmed available, thereby improving the success rate while managing complexity through structured pre-processing
Solution Approach 2:
The patent creates instance copies that mirror the resource requirements of actual pods. These instance copies are used in simulation scheduling to predict resource availability and determine target nodes without affecting actual system state, allowing complex validation to be performed on replicas rather than production resources
2Reliability
If simulation scheduling is performed to determine target nodes and reserve resources, then resource scheduling success rate is improved, but computational overhead increases
Solution Approach 1:
The patent performs simulation scheduling on instance copies rather than full pod deployments. This partial action approach validates resource availability and determines target nodes without committing actual resources or triggering full scheduling workflows, reducing the time cost while maintaining the benefit of pre-validation
Solution Approach 2:
The patent uses temporary instance objects for simulation purposes. These instances are created briefly to perform resource validation and then discarded, avoiding the overhead of actual pod lifecycle management while still providing accurate resource requirement validation
Data Source
AI summary
The embodiments of the present disclosure provide a resource processing method, a resource scheduling method, a computing device and a computer storage medium, where the resource processing method includes: creating, according to the number of a pod needed by a target application, at least one instance corresponding to the number of the pod for the target application; performing, according to a resource requirement parameter of the at least one instance and respective available resources of multiple nodes respectively, simulation scheduling in the multiple nodes to determine at least one target node and a respective reserved resource thereof; and generating, according to a simulation scheduling result, a resource reservation result of the target application, where the resource reservation result is used to instruct to perform resource scheduling for the target application by using the a reserved resource of the at least one target node.