Machine Learning Workload Management Across Edge Computing Sites
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing edge computing systems face challenges in efficiently managing workloads across edge devices, leading to overburdened resources and performance issues due to the inability to effectively distribute tasks and services between edge computing sites.
Innovation Solution
An edge infrastructure management platform utilizing machine learning algorithms to analyze current workloads, predict future demands, and optimize task distribution by transferring tasks between edge devices and fog devices to balance resource usage and maintain performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If tasks are concentrated on centralized locations, then data processing can be performed, but network bandwidth is consumed and latency increases
Solution Approach 1:
The patent segments the centralized data processing function into distributed edge computing nodes deployed at the network periphery. Each edge device independently processes data locally, eliminating the need to transmit raw data to centralized locations, thereby reducing network bandwidth consumption and latency while improving network reliability
Solution Approach 2:
The patent introduces a spatial dimension to data processing by deploying edge computing devices across multiple geographic locations rather than concentrating processing in a single centralized location. This dimensional distribution enables local processing near data sources, reducing transmission distance and time
2Speed
If edge computing devices are deployed to reduce latency, then real-time processing improves, but workload management and resource distribution become challenging
Solution Approach 1:
The patent implements a workload management system that continuously monitors resource utilization metrics across edge devices and uses this feedback to dynamically adjust task allocation. The system collects data on CPU usage, memory availability, and current workload, then redistributes tasks to maintain balanced resource utilization and prevent any single edge device from becoming overwhelmed
Solution Approach 2:
The patent enables edge devices to autonomously manage their own workloads through self-service mechanisms. Each edge device can independently accept, process, and reject tasks based on its current resource capacity, reducing the need for complex centralized coordination while maintaining efficient workload distribution across the edge computing network
3Productivity
If tasks are distributed across multiple edge devices, then resource utilization improves, but system complexity increases
Solution Approach 1:
The patent creates a universal task management platform that can handle multiple types of computing tasks across diverse edge devices through a common interface and standardized protocols. This multi-functional approach allows the same system architecture to manage various workloads (data processing, analytics, inference) on different device types without requiring device-specific management logic, thereby improving resource utilization while controlling system complexity
Data Source
AI summary
A method comprises receiving data corresponding to operation of a plurality of edge devices from respective ones of a plurality of edge computing sites. The data comprises requests received by the edge devices to perform a plurality of tasks. The data is analyzed using a first machine learning algorithm to determine workloads of respective ones of the edge devices. The method further comprises predicting future workloads of the edge devices. The predicting is performed using a second machine learning algorithm and is based on the determined workloads of the edge devices. A determination is made whether to transfer at least a portion of one or more of the tasks from a first edge device to a second edge device. The first and second edge devices are located at first and second edge computing sites, respectively, and the determination is based on one or more of the predicted future workloads.