ML-Based Network Resource Recommendation for Proactive Capacity Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large-scale networks like 4G and 5G wireless networks face challenges in determining when to increase network resources in access and core networks to maintain performance without causing unnecessary costs or disruptions, as existing methods often wait until performance degradation is noticeable to users.
Innovation Solution
A system that collects resource utilization information from network devices and uses a machine learning model trained with historical data to recommend resource increases, allowing for proactive adjustments before performance degradation occurs, utilizing virtual network functions and simulation environments for testing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If network resources are increased proactively using machine learning predictions, then network performance is maintained and user satisfaction is improved, but system complexity and implementation costs increase
Solution Approach 1:
The machine learning model performs preliminary analysis of historical resource utilization data to predict future resource needs before performance degradation occurs. This allows proactive resource allocation decisions to be made in advance, maintaining network performance without waiting for problems to manifest.
Solution Approach 2:
A simulation environment acts as an intermediary between the ML model predictions and actual network deployment. The simulation validates predicted resource configurations before implementation, reducing the risk of introducing complexity-related errors while maintaining reliability.
2Reliability
If network resources are increased earlier using predictive analytics, then performance degradation is prevented, but unnecessary resource costs may increase
Solution Approach 1:
The system dynamically adjusts resource allocation parameters based on ML-predicted utilization patterns rather than using fixed thresholds. This allows resources to be allocated precisely when and where needed, preventing both performance degradation and unnecessary resource consumption.
Solution Approach 2:
The system continuously monitors actual resource utilization and compares it against ML predictions, using this feedback to refine future predictions and adjust resource allocation decisions. This closed-loop approach ensures resources are allocated efficiently without over-provisioning.
3Ease of operation
If traditional threshold-based monitoring is used to detect resource needs, then implementation is simple, but performance degradation is only detected after it occurs
Solution Approach 1:
The patent replaces traditional mechanical threshold-based monitoring with a machine learning-based predictive system. Instead of reacting to fixed thresholds, the ML model analyzes patterns in historical data to predict future resource needs, enabling proactive rather than reactive resource management.
Data Source
AI summary
Resource utilization information including one or more of central processor utilization, memory utilization, and device load values are received from each of a plurality of network devices that compose an access network that is communicatively coupled to a core network. The resource utilization information is provided to a machine learning model to receive an output, wherein the machine learning model has been trained with historical resource utilization information associated with network devices that compose an access network, predetermined thresholds, and an impact to the access network of network resources being added to a core network. A recommended increase in network resources of the core network or the access network is sent to a destination based on the output.