Collaborative Cloud Resource Scheduling via Filter Chains

The patent segments the scheduling function by maintaining separate schedulers for each IaaS service (compute, networking, storage) rather than using a single common scheduler. Each service scheduler operates independently with its own filtering and selection logic, eliminating the need for extensive inter-service control traffic and state sharing while still enabling coordinated resource allocation through service-level independence.

The patent introduces resource candidates as an intermediary mechanism that enables coordination between service schedulers. Each scheduler maintains a local set of resource candidates that can be shared or referenced by other services, allowing joint scheduling decisions to be made through this intermediary structure without requiring direct complex communication between schedulers.

The patent segments the scheduling hierarchy into peer service schedulers that operate simultaneously rather than in a serialized hierarchical manner. Each service scheduler independently filters and selects resources from local resource candidates, enabling parallel processing and eliminating the bottleneck of serialized invocation while maintaining coordination through shared resource candidate structures.

The patent allows each service scheduler to perform partial scheduling actions independently by filtering resource candidates according to service-specific criteria. This partial action approach enables concurrent scheduling operations across multiple services without requiring a single centralized scheduler to perform all actions sequentially, thereby improving overall scheduling efficiency.

The patent makes resource candidates universal by maintaining them as shared data structures that can be accessed and filtered by multiple service schedulers. This allows service-independent schedulers to operate with their own filtering logic while simultaneously improving resource utilization through shared access to the same resource candidate pool, enabling both service independence and efficient resource allocation.

The patent implements feedback mechanisms where service schedulers monitor resource allocation outcomes and adjust their filtering criteria accordingly. This feedback loop enables independent schedulers to learn from each other's actions and make better scheduling decisions, improving overall resource utilization while maintaining service independence through localized decision-making.