Traffic Multiplexing Timestamp Queue Fairness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing queue management techniques in network nodes, such as round-robin scheduling and Weighted Fair Queuing (WFQ), fail to provide fair packet forwarding during fluctuations in network traffic, leading to disproportionate packet drops, especially when traffic patterns are unknown or change over long intervals.
Innovation Solution
A fairness procedure that associates a time-stamp with each packet in a managed group of queues, ensuring that the packet with the earliest time-stamp is forwarded first, thereby proportionally dropping packets from each non-empty queue during congestion, allowing for dynamic adjustment to transmission rate fluctuations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If round-robin scheduling is used to manage multiple queues, then forwarding fairness to each queue is improved, but disproportionate packet drop occurs when traffic patterns fluctuate
Solution Approach 1:
The patent applies dynamics by making the queue weight dynamic rather than static. Instead of predetermined weights, the system continuously adjusts queue weights based on actual traffic rates measured over time intervals. This allows the scheduling mechanism to adapt to fluctuating traffic patterns, ensuring that queues with higher traffic rates receive proportionally more bandwidth while preventing disproportionate packet drops during congestion.
Solution Approach 2:
The patent implements feedback by measuring actual traffic rates from each queue and using this information to adjust scheduling weights. The system monitors packet counts or byte counts over time intervals, compares actual rates against target rates, and dynamically modifies queue weights to achieve proportional fairness. This closed-loop feedback mechanism ensures that forwarding fairness is maintained even when traffic patterns change unexpectedly.
2Loss of substance
If predetermined weighting in WFQ is used to compensate for disproportionate packet drop, then packet drop fairness is improved, but fairness deteriorates when traffic rates fluctuate over long intervals
Solution Approach 1:
The patent resolves this contradiction by making queue weights dynamic rather than predetermined. The system continuously measures actual traffic rates and adjusts weights in real-time, allowing it to adapt to both short-term and long-term traffic fluctuations. This dynamic approach maintains packet drop fairness while simultaneously providing adaptability to changing traffic patterns, eliminating the trade-off present in static WFQ implementations.
Solution Approach 2:
The patent changes the parameter of queue weight from a fixed predetermined value to a dynamically adjusted value based on measured traffic rates. By continuously modifying the weight parameter in response to actual traffic conditions, the system achieves both packet drop fairness and adaptability to traffic fluctuations, resolving the contradiction between these two requirements.
3Ease of manufacture
If a single FIFO queue is used instead of multiple queues, then implementation cost is reduced, but memory device write speed requirements increase
Solution Approach 1:
The patent applies segmentation by dividing the single FIFO queue into multiple separate queues, each handling traffic from a specific input or traffic class. This segmentation allows each queue to be managed independently with its own weight and scheduling priority. The memory device only needs to write at the maximum rate of a single input rather than the sum of all inputs, reducing the write speed requirement while maintaining implementation feasibility.
Solution Approach 2:
By segmenting traffic into multiple queues rather than using a single FIFO queue, the patent reduces the memory write speed requirement from the sum of all input rates to the maximum rate of individual inputs. This segmentation makes the system more implementable with cost-effective memory devices while still enabling sophisticated traffic management and proportional fairness through dynamic weight adjustment.
Data Source
AI summary
Multiplexed traffic in a system where the maximum rate of all inputs exceeds the maximum rate of the output is processed via a separate queue for each input, and at any given forwarding clock cycle the earliest enqueued packet is forwarded via the output. In the event of congestion, a proportionally equal number of packets are dropped from each queue, where proportional equality corresponds to the number of packets dropped per number of packets enqueued. One implementation associates a time-stamp with each enqueued packet to indicate the time of enqueing relative to other enqueued packets. At any given forwarding clock cycle, the packet with the earliest time-stamp is forwarded.