Hybrid Cloud Scheduling With Global Conflict Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud computing systems face inefficiencies in resource allocation due to scheduling conflicts among multiple scheduler instances, leading to unfairness, underutilization of resources, and slow scheduling performance, particularly when handling heterogeneous runtime instances.
Innovation Solution
Implement a hybrid centralized distributive scheduling system with a global resource manager (GRM) that coordinates high-level resource management and conflict resolution across scheduler instances, supplemented by local resource managers (LRMs) on each physical host for detailed scheduling and optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple scheduler instances independently allocate resources to physical hosts, then resource allocation flexibility increases, but scheduling conflicts and resource underutilization occur
Solution Approach 1:
The patent introduces a central scheduler as an intermediary between multiple scheduler instances and physical hosts. This central coordinator receives resource allocation requests from distributed scheduler instances, resolves conflicts by selecting appropriate physical hosts, and returns decisions to the originating schedulers. This mediator architecture enables multiple schedulers to operate independently while maintaining global coordination, thus preserving allocation flexibility while improving scheduling efficiency and reducing conflicts.
2Device complexity
If available resources are offered to one central scheduler instance at a time, then resource allocation simplicity increases, but scheduling performance becomes slow when idle resources exist
Solution Approach 1:
The patent segments the scheduling process into two independent parts: a central scheduler that maintains the global resource pool and handles conflict resolution, and multiple distributed scheduler instances that independently evaluate requests and propose allocations. This segmentation allows parallel processing where multiple schedulers can simultaneously work on different resource requests without interfering with each other, significantly improving scheduling performance while maintaining the simplicity of centralized resource tracking.
Solution Approach 2:
Distributed scheduler instances perform preliminary actions by independently evaluating resource requests, filtering candidate physical hosts, and preparing allocation proposals before submitting them to the central scheduler. This preliminary processing distributes the computational workload and enables parallel evaluation of multiple requests, improving overall scheduling speed while the central scheduler focuses on final conflict resolution and decision-making.
3Measurement precision
If a request-and-reject process is used for resource allocation, then resource allocation accuracy improves, but scheduling efficiency decreases for heterogeneous runtime instances
Solution Approach 1:
Distributed scheduler instances perform preliminary filtering and evaluation of candidate physical hosts based on resource requirements and host capabilities before submitting requests to the central scheduler. This preliminary action pre-resolves many compatibility issues and reduces the number of rejection cycles needed, maintaining allocation accuracy while significantly improving efficiency for heterogeneous runtime instances such as virtual machines, containers, and bare-metal workloads.
Data Source
AI summary
Systems and systems for hybrid centralized distributive scheduling and conflict resolution of multiple scheduler instances that share physical resources in a cloud computing system. The cloud computing system includes a plurality of scheduler instances, a global resource manager (GRM) for high-level resource management and conflict resolution for the scheduler instances, and a plurality of physical hosts. Each physical host has a respective local resource manager (LRM). The scheduler instances are responsible for initially processing of scheduling and resource allocation for resource requests, and proposing candidate physical hosts (and respective resource allocation) for the resource requests to the GRM. The GRM is responsible for conflict resolution through its general conflict resolvers of filtering, sorting and counting. The GRM decides which physical hosts among the candidate physical hosts will run the runtime instances of the resource requests after resolving conflicts among the scheduler instances.