Auto Scaling Event Classification via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing auto scaling technologies in computing service environments often trigger scaling events erroneously due to internal issues like misconfiguration or programming errors, leading to unnecessary resource allocation and increased costs for customers.
Innovation Solution
The implementation of a machine learning-based system that classifies auto scaling events as either legitimate or erroneous by analyzing feature information related to configuration changes in computing service resource groups, allowing for timely notification and potential rollback of erroneous actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If auto scaling events are triggered based on configuration changes, then resource availability is maintained, but erroneous scaling events occur leading to unnecessary resource allocation
Solution Approach 1:
The system performs preliminary classification of auto scaling events using a machine learning model before actual scaling actions are executed. Configuration events are captured and used as features to predict whether an auto scaling event is legitimate or erroneous, allowing preventive action before unnecessary resource allocation occurs
Solution Approach 2:
A machine learning classification service is introduced as an intermediary between configuration changes and auto scaling actions. This intermediary analyzes configuration events and system state information to determine whether scaling events should be executed, filtering out erroneous triggers while maintaining legitimate ones
2Productivity
If machine learning classification is implemented to identify erroneous events, then resource management efficiency improves, but system complexity increases
Solution Approach 1:
The machine learning model is trained using historical data from the same computing service environment, allowing the system to self-improve its classification accuracy over time. The system automatically captures configuration events and system state information, processes them through the classification service, and uses the results to prevent erroneous scaling without requiring manual intervention
Solution Approach 2:
The system implements a feedback loop where classification results are used to prevent erroneous auto scaling events, and the outcomes of these classifications are fed back to continuously improve the machine learning model. Configuration events and system state information are continuously monitored and used to retrain and refine the classification accuracy
Data Source
AI summary
A technology is described for classifying an auto scaling event. An example method may include capturing information related to a configuration event having an impact on a computing service resource group hosted in a service provider environment. Detecting an auto scaling event associated with the computing service resource group may result in programmatically instantiating or stopping a computing service resource, whereupon the information related to the configuration event may be retrieved and the information related to the configuration event may be input to a machine learning model configured to classify the auto scaling event as a valid auto scaling event or an invalid auto scaling event using the information related to the event.