Cooperative Cloud Resource Scheduling via Allocation Plans
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cloud computing systems face inefficiencies due to sporadic and unplanned interactions between workload schedulers and resource schedulers, leading to performance slowdowns and resource fragmentation, as well as challenges in cooperatively scheduling resources for virtual machines and hypervisor-based container workloads.
Innovation Solution
Implementing a method where workload schedulers submit resource allocation plans to resource schedulers, including attributes like allocation specifications, goals, and time constraints, allowing for coordinated resource allocation and fusion of resources, enabling efficient and predictive scheduling with continuous optimization and resource reuse.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If workload schedulers and resource schedulers interact sporadically and unplanned, then resource allocation responds quickly to individual requests, but performance slows down and resource fragmentation occurs
Solution Approach 1:
The resource allocation plan is submitted in advance to the resource scheduler before workloads need to be scheduled. This preliminary action allows the resource scheduler to pre-allocate and prepare resources, avoiding the need for sporadic interactions during workload execution, thereby improving system performance while maintaining allocation responsiveness
Solution Approach 2:
The resource allocation plan acts as an intermediary mechanism between workload schedulers and resource schedulers. It standardizes the interaction interface, enabling planned and coordinated resource allocation rather than ad-hoc requests, which resolves the contradiction between quick responsiveness and overall system performance
2Adaptability or versatility
If workload schedulers and resource schedulers interact frequently and unplanned, then resource allocation can adapt to changing workload needs, but resource fragmentation increases
Solution Approach 1:
By submitting resource allocation plans in advance, the system establishes a stable resource allocation framework before workloads are scheduled. This preliminary planning reduces the need for frequent unplanned interactions, maintaining resource allocation stability while still allowing adaptations through the structured plan modification process
Solution Approach 2:
The resource allocation plan is designed to be dynamic and modifiable. Workload schedulers can submit new plans or modify existing ones as workload needs change, allowing the system to adapt flexibly while maintaining overall stability through the structured plan-based approach rather than ad-hoc changes
3Productivity
If resources are allocated without a coordinated plan, then individual workload requests are fulfilled quickly, but overall resource utilization efficiency decreases
Solution Approach 1:
The resource allocation plan is submitted in advance to the resource scheduler, enabling coordinated resource allocation across multiple workloads. This preliminary planning allows the resource scheduler to optimize resource distribution and avoid fragmentation, improving overall resource utilization efficiency while maintaining allocation speed through the structured process
Solution Approach 2:
The resource allocation plan enables multiple workload requests to be merged and coordinated through a single planning process. The resource scheduler can allocate resources to multiple workloads in a coordinated manner, improving overall resource utilization efficiency while maintaining allocation speed through the structured plan-based approach
Data Source
AI summary
Apparatus and methods for scheduling computing resources is disclosed that facilitate the cooperation of resource managers in the resource layer and workload schedulers in the workload layer working together so that resource managers can efficiently manage and schedule resources for horizontally and vertically scaling resources on physical hosts shared among workload schedulers to run workloads.