Resource Scheduling Balancing User Instance Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cloud-computing-based resource scheduling fails to balance user instance distributions across hosts, leading to frequent failures due to resource competition and waste, as it does not consider the distribution of user instances with physical hosts, thereby affecting user experience.
Innovation Solution
A method and apparatus for scheduling resources that traverse user instance distributions, detect trigger conditions for adjustment, and migrate instances from overloaded hosts to underutilized ones, optimizing resource allocation and preventing risks by ensuring a balanced distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional resource scheduling places user instances based solely on current resource status, then resource allocation is simple and fast, but user instances become concentrated on few hosts causing frequent failures and service quality degradation
Solution Approach 1:
The patent segments the scheduling decision into two independent parts: resource status evaluation and user instance distribution evaluation. The scheduling system first assesses host resource status, then separately evaluates user instance distribution patterns, and only then makes placement decisions. This segmentation allows the system to consider both resource availability and distribution balance without excessive complexity, preventing instance concentration on few hosts while maintaining manageable scheduling logic.
Solution Approach 2:
The patent implements preliminary evaluation of user instance distribution before actual instance placement. The system pre-assesses the distribution status of user instances across hosts and uses this pre-evaluation information to guide scheduling decisions. This preliminary action ensures that instances are not placed in a way that would cause concentration on few hosts, thereby improving service quality and reliability before problems occur.
2Ease of operation
If user instances are concentrated on few hosts to simplify scheduling, then scheduling operation is easy, but host failures affect all instances and user experience deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the scheduling system continuously monitors and evaluates user instance distribution across hosts. This feedback information about distribution patterns is fed back into the scheduling decision process, allowing the system to automatically adjust placement strategies to prevent concentration. The feedback loop maintains service continuity by ensuring instances are distributed across multiple hosts without requiring complex manual intervention.
Solution Approach 2:
The scheduling system performs self-evaluation of user instance distribution and automatically makes adjustments without external intervention. The system independently assesses whether instances are properly distributed across hosts and autonomously makes placement decisions to maintain balanced distribution. This self-service capability ensures service continuity while keeping scheduling operations simple and automated.
3Productivity
If resource scheduling does not consider user instance distribution, then scheduling process is simple and quick, but host resource usage becomes unbalanced causing waste and failures
Solution Approach 1:
The patent applies partial evaluation by focusing specifically on user instance distribution characteristics rather than comprehensively re-evaluating all host parameters. The system performs a targeted assessment of instance distribution patterns and applies this partial evaluation to scheduling decisions. This approach maintains scheduling efficiency while preventing resource waste through balanced distribution, avoiding the need for complete re-assessment of all system parameters.
Solution Approach 2:
The patent introduces distribution balance as an additional evaluation parameter in the scheduling decision process. By adding this new parameter to the traditional resource status evaluation, the system achieves more balanced host resource usage without significantly increasing complexity. The inclusion of distribution balance as a scheduling parameter prevents resource concentration and waste while maintaining relatively efficient scheduling operations.
Data Source
AI summary
Disclosure method that includes traversing a user instance distribution of user instances of a user on hosts in a cluster to detect whether a trigger condition for user scheduling of adjustment of the user instance distribution is satisfied; migrating the user instance to be migrated out from the one or more hosts from which the user instances are to be migrated out to the one or more second hosts to which the one or more user instances are to be migrated in if the trigger condition for user scheduling is satisfied. The implementations of the present disclosure may schedule resources based on user instance distributions of users on hosts and achieve a balanced distribution of user instances. The implementations further avoid placing user instances of all users into a small number of hosts, increase the ability to prevent risks, and enhance the user experience.