MEC Policy Optimization via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Application developers often over- or under-subscribe MEC resources, leading to higher service costs and unnecessary resource constraints due to inaccurate policy settings for minimum latency and instance requirements, resulting in suboptimal performance and resource allocation.
Innovation Solution
A machine-learning module analyzes network performance data to generate optimized policy suggestions for MEC services, allowing developers to select cost-effective settings while optimizing resource allocation, with a central network device refining policies across multiple edge clusters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If developers manually set policy parameters for MEC services, then they can define minimum latency and instance requirements, but they often over- or under-subscribe resources leading to higher service costs and unnecessary resource constraints
Solution Approach 1:
The system enables self-service through automated policy optimization. The machine learning module autonomously analyzes network performance data and generates optimized policy recommendations without requiring manual intervention from developers. This self-optimizing mechanism continuously improves policy accuracy while reducing resource waste, as the system automatically adjusts latency and instance requirements based on actual network conditions and application performance metrics.
2Reliability
If developers over-subscribe MEC resources to ensure performance requirements are met, then application performance can be maintained, but service costs increase and resource constraints become unnecessary
Solution Approach 1:
The system dynamically changes policy parameters including minimum latency requirements and minimum instance requirements based on analyzed network performance data. By adjusting these parameters optimally rather than using conservative over-provisioning, the system maintains application performance reliability while reducing the quantity of subscribed resources, thereby lowering service costs and eliminating unnecessary resource constraints.
3Quantity of substance
If developers under-subscribe MEC resources to reduce service costs, then service costs decrease, but performance requirements cannot be met leading to suboptimal application performance
Solution Approach 1:
The system implements feedback mechanisms where network performance data is continuously collected, analyzed, and used to generate optimized policy recommendations. This feedback loop ensures that resource subscription levels are optimally matched to actual performance requirements, preventing both over-provisioning and under-provisioning. The automated analysis of performance metrics enables the system to identify the minimum necessary resources to meet application performance requirements, thereby reducing service costs without compromising reliability.
4Ease of operation
If manual policy configuration is used for MEC services, then developers have control over service parameters, but inaccurate settings result in suboptimal performance and resource allocation
Solution Approach 1:
The machine learning module acts as an intermediary between manual policy configuration and actual resource allocation. It receives developer-defined service parameters, analyzes network performance data, and generates optimized policy recommendations that bridge the gap between simple manual configuration and optimal resource allocation. This intermediary processing maintains ease of operation for developers while significantly improving resource allocation efficiency through data-driven optimization.
Data Source
AI summary
Systems and methods provide a MEC policy optimization service. A network device applies, in a first edge cluster of the application service layer network, a policy for an application service that supports a customer application; receives network performance data related to execution of the application service; identifies, based on the network performance data, an improved policy to optimize the first policy for the customer application in the first edge cluster; and sends a model of the improved policy to a central network device for the application service layer network.