Automatic Resource Scaling With Future-Load Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current auto scaling methods in computing and storage resources are inaccurate due to unknown service changes and delays, leading to resource waste and inefficiency in public clouds, edge clouds, and content delivery networks.
Innovation Solution
An automatic resource scaling method that utilizes both current and future load information to determine whether to perform scale-out or scale-in, considering conditions and steps based on historical data and resource configuration to optimize resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If resource scaling is performed based only on current resource utilization, then the system can respond quickly to current load, but the scaling decisions are inaccurate due to unknown service changes and delays
Solution Approach 1:
The system performs preliminary actions by collecting and analyzing historical load data before making scaling decisions. It uses past load patterns to predict future load, allowing the system to prepare scaling actions in advance based on predicted trends rather than waiting for current thresholds to be triggered, thus reducing scaling delays while improving accuracy
Solution Approach 2:
The system implements feedback mechanisms by continuously monitoring current resource utilization and comparing it with predicted future load based on historical data. This feedback loop allows the system to adjust scaling decisions dynamically, ensuring both timely responses and accurate predictions by validating predictions against actual observed patterns
2Reliability
If resources are scaled out based on current high load, then service requirements are met, but resources are wasted when load decreases significantly after scaling
Solution Approach 1:
The system performs preliminary analysis of historical load patterns before executing scaling-out actions. By predicting whether high current load is part of a sustained trend or a temporary spike, the system can make more informed decisions about whether to scale out, thereby avoiding resource waste from premature or unnecessary scaling while ensuring service requirements are met when scaling is truly needed
3Loss of energy
If resources are scaled in based on current low load, then resource utilization is improved, but service quality is impacted when load increases after scaling
Solution Approach 1:
The system performs preliminary prediction using historical load data before executing scaling-in actions. By analyzing whether current low load is a sustained pattern or a temporary dip, the system can determine the appropriate timing for scaling in, thereby improving resource utilization efficiency while avoiding service quality degradation that would result from premature scaling-in during temporary load reductions
Data Source
AI summary
An automatic resource scaling method includes that first load information and second load information are obtained, where the first load information indicates current actual load information, and the second load information is load information used to estimate a future load; and whether to perform resource scale-out or resource scale-in is determined based on the first load information and the second load information. A control node can determine a current actual load status based on the first load information, and can further estimate a load in a future period of time based on the second load information, to determine, based on the current actual load status and the estimated load status in the future period of time, whether to perform resource scale-out or scale-in.