Computing Power Network Scheduling with Multi-Objective Platform Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for collaborative scheduling of wide-area computing resources fail to adapt to complex and changing network environments, often assigning tasks to suboptimal network paths, leading to inefficiencies and performance issues due to variability and resource heterogeneity.
Innovation Solution
A method and system for computing power network scheduling that filters and combines storage and computing power platforms based on comprehensive weight decision of network performance indicators, using a multi-objective optimization function to determine the optimal combination for task deployment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If simple task assignment policies are used, then the scheduling system is easy to operate, but task allocation efficiency and network performance deteriorate due to ignoring network path variability and resource heterogeneity
Solution Approach 1:
The patent transforms static task assignment into dynamic scheduling by continuously monitoring and adjusting network performance parameters (bandwidth, delay, packet loss) and resource parameters (CPU, memory, storage) in real-time. The scheduling algorithm adapts to changing conditions by re-evaluating task assignments based on current network state, thereby maintaining high allocation efficiency while managing system complexity through automated parameter-driven decisions
Solution Approach 2:
The system implements comprehensive feedback mechanisms that monitor network path performance and resource utilization status, feeding this information back to the scheduling algorithm. The scheduling decision-making process incorporates real-time feedback about network conditions and resource availability, enabling continuous optimization of task allocation to match actual system state and improve overall productivity
2Device complexity
If static policies are used for task allocation, then the system complexity is low, but adaptability to complex and changing network environments deteriorates
Solution Approach 1:
The patent implements dynamic scheduling policies that adapt to changing network conditions and resource availability in real-time. The system continuously updates its scheduling decisions based on monitored network performance metrics and resource status, transitioning from static to dynamic task allocation. This enables the system to adapt to network environment changes while managing complexity through structured dynamic adjustment mechanisms
Solution Approach 2:
The system performs preliminary actions by pre-evaluating multiple potential task assignments and preparing alternative scheduling plans before actual task execution. The scheduling algorithm anticipates potential network conditions and resource constraints, pre-calculating optimal assignments that can be quickly deployed when needed, thereby enhancing adaptability without proportionally increasing operational complexity
3Device complexity
If tasks are assigned to network paths with poor performance, then the scheduling decision is simple, but transmission delay and bandwidth bottlenecks increase
Solution Approach 1:
The patent replaces simple mechanical task assignment rules with an intelligent scheduling system that uses network performance modeling and optimization algorithms. Instead of static routing, the system employs computational methods to analyze network conditions, predict performance metrics, and select optimal paths automatically, substituting complex decision-making with automated intelligent algorithms that minimize transmission delay
4Productivity
If comprehensive weight decision of network performance is implemented, then computational resource utilization and task scheduling efficiency improve, but the scheduling system complexity increases
Solution Approach 1:
The patent segments the comprehensive scheduling problem into manageable sub-problems by separating different decision-making aspects: network path selection, resource allocation, and task scheduling. The system divides the complex weight decision process into multiple evaluation criteria (bandwidth, delay, packet loss, resource availability) that can be independently monitored and optimized, reducing overall system complexity through modular decision architecture
Data Source
AI summary
Method for computing power network scheduling service based on comprehensive weight decision of network performance and the system thereof are provided. A computing power network service platform combines the filtered storage platforms and at least one filtered computing power platform that do not belong to a same data center to obtain a plurality of combinations; constructs a multi-objective optimization function under a storage-computing separation scenario based on the different network performance indicators, subjective weight vectors of the different network performance indicators, objective weight vectors of the different network performance indicators between the storage platforms and the computing power platforms for each combination, and resource demand of the user, and solves the multi-objective optimization function to obtain the optimal combination of the storage platform and the computing power platform that meets an user demand; and deploys job and computational data of the user to the optimal combination for storage and computation.