Communication System Scale-Out via Dynamic Threshold Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems face challenges in scaling out elements timely and accurately due to low prediction accuracy of network load, leading to potential over-scaling or under-scaling.
Innovation Solution
A scale-out execution system that predicts network load based on performance index values before a reference time point, determines whether the predicted load exceeds a threshold value that increases with prediction time, and executes scale-out when the predicted value exceeds the threshold at one or more prediction time points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If scale-out is executed early based on low-accuracy prediction to have a safety margin, then the network load increase risk is reduced, but resource waste increases due to unnecessary scale-out
Solution Approach 1:
The patent changes the threshold parameter dynamically based on prediction accuracy. When prediction accuracy is low, a higher threshold is applied, requiring stronger evidence of network load increase before triggering scale-out. This resolves the contradiction by adjusting the decision criterion (threshold) according to the reliability of the prediction, preventing unnecessary scale-out while maintaining safety margins.
Solution Approach 2:
The system dynamically adjusts the threshold value based on the predicted accuracy level. Instead of using a fixed threshold, the threshold becomes a variable that adapts to prediction quality. This dynamic adjustment allows the system to be more conservative when predictions are uncertain and more aggressive when predictions are reliable, resolving the contradiction between safety and resource efficiency.
2Loss of substance
If scale-out is delayed to wait for higher accuracy prediction, then resource efficiency improves, but network load may increase beyond acceptable levels
Solution Approach 1:
The patent applies different threshold parameters depending on the prediction time horizon and accuracy. For longer prediction periods with lower accuracy, higher thresholds are used to avoid premature scale-out, while for shorter periods with higher accuracy, lower thresholds enable timely scale-out. This parameter adaptation resolves the contradiction between waiting for accuracy and controlling network load.
Solution Approach 2:
The system performs preliminary scale-out actions when predictions indicate future network load increases, even before the actual peak occurs. By using multiple prediction time points and comparing against adaptive thresholds, the system can trigger scale-out in advance while maintaining resource efficiency, resolving the contradiction between proactive action and resource waste.
3Measurement precision
If multiple prediction time points are used to improve decision accuracy, then the scale-out timing becomes more precise, but the system complexity increases
Solution Approach 1:
The patent segments the prediction period into multiple discrete time points, evaluating network load predictions at each segment. This segmentation allows the system to identify the optimal scale-out timing by comparing predictions across different future moments, improving timing precision while keeping the analysis structured and manageable.
Solution Approach 2:
The system performs predictions at multiple time points (excessive action) but only triggers scale-out when specific conditions are met at one or more of these points. This partial execution approach - performing more predictions than minimally required but selectively acting - improves timing precision without fully committing to the complexity of continuously acting on all prediction results.
Data Source
AI summary
A communication system is appropriately scaled out. An AI predicts, based on a performance index value before a given reference time point, a network load at a plurality of prediction time points included in a prediction period from the given reference time point until a predetermined time after the given reference time point. A policy manager determines, for each of the plurality of prediction time points, whether a predicted value of the network load at the prediction time point exceeds a threshold value, which increases as a length of time from the reference time point until the prediction time point increases. The policy manager, a life cycle manager, a container manager, and a configuration manager execute scale-out of an element included in the communication system when it is determined that the predicted value exceeds the threshold value at a part or all of the plurality of prediction time points.