Variable Rate Token Bucket for Packet Marking Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing packet marking methods in networks, such as Hierarchical Token Bucket (HTB) and Rainbow Fair Queuing (RFQ), face inefficiencies due to high computational and storage complexity, and slow adaptation to changing network conditions, especially in transient situations with significant fluctuations in traffic arrival rates.
Innovation Solution
A variable rate token bucket system at Edge Nodes that estimates packet arrival rates and generates packet values using a Throughput Value Function, allowing for quick adaptation to network changes with reduced computational complexity, emulating HTB behavior while maintaining only three state variables.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Hierarchical Token Bucket (HTB) based packet marking is used to achieve optimal bandwidth sharing control, then bandwidth sharing accuracy is improved, but computational and storage complexity increases making it impractical for real applications
Solution Approach 1:
The patent replaces the complex HTB mechanism with a simplified token bucket approach that uses lightweight, easily maintainable state variables. Instead of managing hierarchical token buckets with multiple levels and complex state transitions, the invention uses a single token bucket with a finite state machine that tracks only essential states (idle, sending, paused, finished), dramatically reducing computational and storage requirements while maintaining practical implementability
Solution Approach 2:
The patent extracts and removes the hierarchical structure from the token bucket mechanism, keeping only the essential single-level token bucket functionality needed for packet marking. By taking out the complex hierarchical layers and retaining only the core token bucket operation combined with a simple finite state machine, the solution achieves practical complexity reduction while preserving the essential bandwidth sharing control capability
2Stability of the object's composition
If Exponentially Weighted Moving Average (EWMA) scheme is used in Rainbow Fair Queuing to estimate arrival rate, then stability for steady traffic loads is improved, but adaptation speed to transient situations deteriorates resulting in long stabilization periods
Solution Approach 1:
The patent implements a dynamic adaptation mechanism where the token bucket parameters (bucket threshold and filling rate) are not fixed but can be adjusted based on observed traffic conditions. The finite state machine transitions between states based on arrival rate changes, allowing the system to dynamically adapt its behavior. This dynamic approach enables fast response to transient situations while maintaining stability during steady-state operation, overcoming the limitation of static EWMA schemes
Solution Approach 2:
The patent incorporates feedback mechanisms where the system continuously monitors packet arrivals and uses this information to adjust token bucket parameters and state transitions. The finite state machine uses feedback from observed traffic patterns to determine when to transition between states (idle, sending, paused, finished), enabling the system to adapt quickly to changing conditions while maintaining stability, thus resolving the contradiction between stability and adaptability
Data Source
Figure 1
Figure 2
Figure 3
AI summary
At an Edge Node, a method of handling data packets in order to mark the packets with respective packet values indicative of a level of importance. The method comprises implementing a variable rate token bucket to determine an estimated arrival rate of a flow of packets. The method comprises receiving a data packet, updating the estimated arrival rate to an updated arrival rate based on a token level of the token bucket and generating a random or pseudo-random number within a range with a limit determined by the updated arrival rate. The method further comprises identifying an operator policy which determines a level of service associated with the flow of packets, and a Throughput Value Function (TVF) associated with said policy, and then applying the TVF to the random number to calculate a packet value. The packet value is included in a header of the packet.