Flow Admission Control Using Experiential Capacity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network operators face challenges in ensuring acceptable Quality of Experience (QoE) for users due to complex relationships between Quality of Service (QoS) metrics and QoE, especially with multiple data flows and varying link conditions, making it difficult to determine which QoS metrics to focus on for optimal user experience.
Innovation Solution
The introduction of a machine-learning based system that determines the 'experiential capacity' of a network, which assesses the ability to carry traffic while maintaining acceptable QoE, by graphing network states as traffic matrices and using a machine-learning engine to classify acceptable network states, allowing for easier decision-making on admitting or rejecting new data flows based on whether the resulting network state falls within these acceptable boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If network operators monitor and manage multiple QoS metrics to ensure acceptable QoE, then the quality of user experience is improved, but the complexity of network management increases
Solution Approach 1:
The patent segments the complex QoE management problem into distinct network states (acceptable and unacceptable) based on traffic matrix patterns. By classifying network states rather than managing individual QoS metrics, the system simplifies decision-making while maintaining QoE guarantees.
Solution Approach 2:
The patent transforms the management approach from monitoring multiple continuous QoS metrics to classifying discrete network states based on traffic matrix characteristics. This parameter transformation simplifies the management complexity while preserving the ability to ensure acceptable QoE.
2Measurement precision
If network operators explicitly measure and evaluate each possible network state to ensure acceptable QoE, then the accuracy of QoE assessment is improved, but the measurement and processing complexity increases
Solution Approach 1:
The patent performs preliminary classification of network states by training a machine learning model on historical traffic matrices to identify patterns associated with acceptable and unacceptable QoE. This preliminary action enables efficient real-time decision-making without requiring exhaustive measurement of each possible state.
Solution Approach 2:
The patent uses traffic matrices as simplified representations (copies) of complex network states. By analyzing these matrix representations rather than measuring all underlying QoS metrics explicitly, the system achieves accurate QoE assessment with reduced measurement complexity.
3Productivity
If the network admits more data flows to increase capacity utilization, then the productivity of the network is improved, but the ability to maintain acceptable QoE deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the machine learning model continuously classifies network states based on current traffic matrices. This feedback enables dynamic flow admission decisions that maintain acceptable QoE while maximizing capacity utilization, as the system can identify when the network is approaching unacceptable states.
Data Source
AI summary
An example method of operating a network may include determining whether a flow is to be added to the network based on: a flow type of the flow, a link condition of the flow, and for each possible combination of flow type and link condition out of multiple flow types and multiple link conditions, the number of flows currently carried on the network that correspond to the respective combination.