On-Device Resource Scheduling for Flexible AI Model Deployment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Intelligent application models deployed on devices with limited system resources face issues such as poor deployment flexibility, low overall operating efficiency, and affect the operation of existing functional modules due to limited resources, leading to high latency, poor flexibility, difficult management, and high data security risks.
Innovation Solution
A resource scheduling method that adaptively schedules system resources, utilizing idle resources to invoke intelligent application models, improving resource utilization and deployment flexibility without affecting existing functional modules, by creating tasks based on current resource usage and intelligent application instances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If intelligent application models are deployed on a device with limited system resources, then the deployment speed is improved, but the deployment flexibility deteriorates
Solution Approach 1:
The patent implements dynamic resource scheduling that adapts to changing system conditions. The scheduling system continuously monitors resource availability and dynamically adjusts the deployment of intelligent application models, allowing the system to transition between different operational states based on real-time resource conditions, thereby maintaining both fast deployment and flexibility.
Solution Approach 2:
The system changes operational parameters by adjusting resource allocation strategies based on system state. When resources are abundant, more models can be deployed; when resources are constrained, the system prioritizes critical models. This parameter-based adaptation resolves the contradiction between rapid deployment and deployment flexibility.
2Loss of time
If intelligent application models are deployed on a device with limited system resources, then the latency is reduced, but the overall operating efficiency deteriorates
Solution Approach 1:
The system applies partial action by selectively deploying only the necessary intelligent application models based on current task requirements and resource availability. Instead of deploying all possible models, the system deploys a subset that is sufficient for current operations, reducing latency while maintaining overall system efficiency through targeted model selection and execution.
3Power
If system resources are occupied by intelligent application models, then the processing capability is improved, but the operation of existing functional modules deteriorates
Solution Approach 1:
The patent segments system resources into dedicated portions for intelligent application models and existing functional modules. The resource scheduling system divides available resources and allocates them dynamically, ensuring that intelligent models receive necessary computational power while existing functional modules retain sufficient resources to maintain stable operation, thus resolving the contradiction between enhanced processing capability and operational reliability.
Data Source
AI summary
Disclosed are a resource scheduling method, an electronic device and a storage medium. The resource scheduling method may include: acquiring an intelligent application processing request acquiring current resource usage information; matching an intelligent application instance according to the intelligent application processing request and creating a task according to the resource usage information and the intelligent application instance, to process the intelligent application processing request.