Random Noise Queue Thresholds for Port Fairness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In network devices, especially multi-port switches, incoming traffic often leads to oversubscription of output queues, resulting in packet drops and unfairness across ports due to synchronized arrival and departure processes, which can be costly to address with per-source, per-destination, per-class-of-service admission control, especially when the number of input ports is large.
Innovation Solution
Implementing a random adjustment of queue thresholds in output buffers to determine whether to enqueue packets or flow-control ports, using a network processor ASIC with programmable logic to generate random numbers and adjust thresholds dynamically, thereby preventing queue overflow and ensuring fairness across ports.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If per-source, per-destination, per-class-of-service admission control is implemented to ensure fairness, then fairness across ports is improved, but device complexity and processing overhead increase significantly
Solution Approach 1:
The patent changes the parameter from detailed per-source, per-destination, per-class-of-service admission control to a simpler queue length threshold-based control. Instead of tracking multiple dimensions of traffic flow, the system monitors only the queue length and applies a randomly adjusted threshold, dramatically reducing processing complexity while maintaining fairness through the randomization mechanism
Solution Approach 2:
The patent replaces expensive, long-term state tracking (per-source, per-destination, per-class admission counts) with a cheap, instantaneous queue length check against a randomly generated threshold. Each decision is made independently based on current queue state rather than accumulated historical data, reducing memory requirements and processing overhead
2Stability of the object's composition
If queue threshold is fixed to ensure predictable behavior, then system stability is improved, but fairness across multiple input ports deteriorates due to synchronized arrivals and departures
Solution Approach 1:
The patent transforms the static queue threshold into a dynamic parameter that is randomly adjusted for each enqueue operation. Instead of a fixed threshold that causes synchronized behavior, the threshold becomes a random variable that changes with each packet arrival, breaking the synchronization pattern while maintaining system stability through the bounded nature of the random adjustment
Solution Approach 2:
The patent introduces asymmetry into the queue management system by using random threshold adjustment instead of symmetric, deterministic threshold comparison. This asymmetric approach prevents multiple input ports from following identical decision patterns, thereby eliminating the unfairness caused by synchronized arrivals and departures while preserving overall system stability
3Reliability
If random noise is added to queue threshold to improve fairness, then source fairness is improved, but predictability of queue behavior deteriorates
Solution Approach 1:
The patent applies partial randomization rather than complete randomness to the queue threshold. The random adjustment is applied only to the threshold value, not to the fundamental queue management logic, maintaining predictable bounds and behavior while introducing enough variability to achieve fairness. The random component is partial, allowing the system to retain most of its deterministic characteristics
Data Source
AI summary
A network device receives a packet at a port from among a number of ports. A queue threshold is randomly adjusted for a queue having an output buffer to which the packet is stored. When a fullness of the queue exceeds the adjusted queue threshold, at a time when the packet is to be enqueued into the queue, an action is performed. The action includes either dropping the received packet, or flow-controlling the port at which the packet is received.