API Endpoint Selection Using Reinforcement Learning for QoE Assurance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques fail to provide a learning-based approach for continuously improving microservice and API service endpoint selections in dynamic environments, leading to unpredictable quality-of-experience (QoE) assurance in large hybrid multicloud settings.
Innovation Solution
Implementing adaptive reinforcement learning to train a machine learning model using quality-of-experience (QoE) data for endpoint selection, enabling self-managed algorithms that optimize endpoint choices based on historical data and QoE history.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional techniques are used for endpoint selection, then the system is simple and easy to implement, but the quality-of-experience assurance is unpredictable and cannot continuously improve
Solution Approach 1:
The system employs self-managed adaptive reinforcement learning algorithms that automatically train and improve endpoint selection models without human intervention. The machine learning model continuously learns from historical QoE data and makes autonomous decisions about optimal endpoint selections, enabling the system to self-optimize its performance over time.
Solution Approach 2:
The system implements feedback mechanisms by collecting quality-of-experience data from historical service requests and using this data to retrain the machine learning model. This closed-loop feedback allows the system to learn from past performance and continuously improve its endpoint selection accuracy, directly addressing the unpredictability of QoE assurance.
2Productivity
If a machine learning model is trained using historical QoE data, then endpoint selection accuracy continuously improves, but the system complexity increases
Solution Approach 1:
The system performs preliminary training of the machine learning model using historical QoE data before deployment. The model is pre-trained on past service request patterns and QoE measurements, enabling it to make accurate endpoint selection decisions from the outset. This preliminary action establishes a foundation for continuous improvement without requiring complex real-time learning during service execution.
Solution Approach 2:
The system dynamically adjusts model parameters and selection probabilities based on incoming QoE data. The reinforcement learning algorithm modifies the endpoint selection vector probabilities according to observed performance metrics, allowing the system to adapt to changing conditions while maintaining a manageable level of complexity through parameter optimization rather than architectural complexity.
3Reliability
If adaptive reinforcement learning is implemented, then the system continuously improves endpoint selections, but the computational resources and training time increase
Solution Approach 1:
The machine learning model is trained in advance using historical QoE data, establishing a pre-optimized endpoint selection strategy. This preliminary training allows the system to make intelligent decisions during service execution without requiring intensive real-time computational resources, as the heavy lifting of pattern recognition and model training is completed beforehand.
Solution Approach 2:
The system applies partial reinforcement learning by updating model parameters periodically based on accumulated QoE data rather than continuously training in real-time. This approach achieves continuous improvement of endpoint selection while significantly reducing instantaneous computational resource consumption, as the model only updates its parameters when new QoE data becomes available for processing.
Data Source
AI summary
A method includes: receiving, by a processor set, a request for a web-based service; generating, by the processor set, a feature vector including values based on parameters of the request; generating, by the processor set, an endpoint selection vector including plural probabilities corresponding to plural endpoints, wherein the endpoint selection vector is generated using the feature vector with a machine learning model; selecting, by the processor set, one of the plural endpoints based on the plural probabilities; and invoking, by the processor set, the selected endpoint.