Multi-Policy Intelligent Scheduling for Heterogeneous Computing Clusters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computing power scheduling policies primarily focus on resource utilization and load balance, neglecting user-specific requirements such as cost, bandwidth usage, and computing time, which limits their effectiveness in meeting diverse computing needs, especially in smart city scenarios.
Innovation Solution
A multi-policy intelligent scheduling method using a Markov decision process model combined with reinforcement learning to set execution policies based on heterogeneity of computing clusters, user requirements, and task characteristics, optimizing task scheduling across various computing clusters to achieve cost-effective and efficient resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional resource-oriented scheduling policies are used to achieve load balance or optimal resource utilization, then resource utilization is improved, but user-specific requirements such as cost, bandwidth usage, and computing time are neglected
Solution Approach 1:
The patent implements dynamic scheduling policies that can adapt to different user requirements and task characteristics. The system uses reinforcement learning to dynamically adjust scheduling decisions based on real-time computing power center states, user preferences, and task properties, transforming the static resource-oriented scheduling into a dynamic user-aware scheduling system.
Solution Approach 2:
The patent changes the scheduling parameters from purely resource-utilization-based metrics to multi-dimensional parameters including user requirements, task characteristics, and computing power center states. By modifying the decision-making parameters to incorporate user-specific factors such as cost sensitivity, bandwidth requirements, and time constraints, the system achieves both high resource utilization and user requirement satisfaction.
2Adaptability or versatility
If reinforcement learning with Markov decision process is used to learn optimal scheduling policies, then user requirement satisfaction is improved, but computational complexity increases
Solution Approach 1:
The patent segments the complex scheduling problem into manageable components using Markov decision processes. By dividing the state space into discrete states and action spaces, and using segmented training approaches, the system makes the reinforcement learning problem computationally tractable while still achieving user-aware scheduling optimization.
Solution Approach 2:
The system uses self-service mechanisms through automated reinforcement learning training and policy optimization. The scheduling system automatically learns optimal policies through interaction with the computing environment, eliminating the need for manual policy configuration and reducing long-term computational overhead despite initial training complexity.
Data Source
AI summary
The present disclosure belongs to the field of intelligent computing technologies, and relates to a multi-policy intelligent scheduling methods and apparatuses oriented to heterogeneous computing power. The method includes: step 1, setting an execution policy of a task based on heterogeneity of computing clusters, differences of computing tasks and a user requirement, and constructing a Markov decision process model by adopting a reinforcement learning method combined with the execution policy; step 2, adopting a proximal policy optimization to solve an optimal task scheduling policy of the task input by the user based on the constructed Markov decision process model; step 3, scheduling the task to a corresponding computing cluster for execution based on the optimal task scheduling policy.