Multi-Timescale Token Bucket for QoS Fairness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network quality of service (QoS) technologies, such as per flow weighted fair queuing, struggle to provide fairness across multiple timescales, especially as the number of flows increases, and existing service level agreements only account for instantaneous traffic behavior, failing to ensure fairness on longer timescales.
Innovation Solution
Determining maximum bucket sizes for token buckets based on drop precedence and timescale to control transmission rates, allowing for fairness across multiple timescales by scaling rate factors and adjusting token rates, and implementing a Multi-Timescale Bandwidth Profile that extends current bandwidth profiles to support quality of service fairness on various timescales.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If per flow weighted fair queuing is used to provide fairness, then fairness is improved, but scalability deteriorates as the number of flows increases
Solution Approach 1:
The patent segments the single fairness mechanism into multiple token buckets operating at different timescales (e.g., short-term, medium-term, long-term buckets). Each bucket handles fairness for specific time intervals, dividing the complex problem of multi-flow fairness into manageable temporal segments that scale better with increasing flow counts
Solution Approach 2:
The patent introduces a temporal dimension to fairness by operating token buckets at multiple timescales simultaneously. This transforms the traditional single-timescale fairness approach into a multi-dimensional fairness system where each timescale addresses different fairness requirements, improving scalability without sacrificing fairness
2Productivity
If service level agreements account for instantaneous traffic behavior, then short-term performance is improved, but fairness on longer timescales deteriorates
Solution Approach 1:
The patent implements periodic token bucket refilling at multiple timescales, where each bucket operates with its own periodic cycle. This allows the system to maintain high instantaneous throughput during active periods while periodically redistributing resources across different timescales to ensure long-term fairness
Solution Approach 2:
The patent creates dynamic resource allocation by allowing token buckets to adapt their behavior across different timescales. The system dynamically adjusts resource distribution based on current traffic conditions while maintaining fairness guarantees over longer periods through coordinated multi-timescale operation
3Reliability
If token bucket size is increased to improve fairness, then resource allocation fairness is improved, but transmission rate control precision deteriorates
Solution Approach 1:
The patent segments the token bucket mechanism into multiple buckets with different sizes and timescales. Each bucket is optimized for its specific timescale, with smaller buckets providing precise short-term control and larger buckets ensuring long-term fairness, thereby maintaining both fairness and control precision simultaneously
Data Source
AI summary
To support quality of service fairness in a communication network, a network node determines a maximum bucket size, for a token bucket controlling a transmission rate from the network node over a transport line between a radio access network and a core network, based on a drop precedence and a given timescale.