Egress Credit Threshold Control for Low-Latency Queue Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing networking systems face inefficiencies in managing egress credits due to high latency and round-trip times, leading to wasted credits and reduced utilization of egress ports, particularly when data flow to an egress port suddenly stops.
Innovation Solution
Optimizing egress credit request thresholds based on round-trip time (RTT) to the egress scheduler, adjusting the queue size thresholds for requesting different credit rates to minimize unused credits and improve port utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the queue size threshold for requesting egress credits is set high, then the egress port utilization is improved, but the latency increases and excess credits are wasted
Solution Approach 1:
The patent implements dynamic adjustment of credit request thresholds based on the round-trip time (RTT) to the egress scheduler. The threshold is calculated as: threshold = RTT × credit_rate. This dynamic approach allows the system to adapt to varying network conditions, optimizing the balance between port utilization and latency. When RTT is high, the threshold increases to prevent premature credit requests that would result in wasted credits, while when RTT is low, the threshold decreases to reduce latency.
Solution Approach 2:
The patent performs preliminary calculation of the optimal credit request threshold using the round-trip time before actually requesting credits. By pre-computing the threshold based on RTT × credit_rate, the system ensures that credit requests are made at the optimal moment, avoiding both premature requests (which waste credits) and delayed requests (which increase latency). This preliminary action resolves the contradiction by proactively determining the optimal request timing.
2Speed
If the credit request rate is increased, then the responsiveness to queue changes is improved, but the number of unused credits increases
Solution Approach 1:
The patent implements a feedback mechanism where the virtual output queue continuously monitors its own queue size and the round-trip time to the egress scheduler. Based on this feedback, the queue dynamically adjusts the credit request threshold using the formula: threshold = RTT × credit_rate. This feedback loop ensures that credit requests are made at the optimal moment, improving responsiveness while preventing the issuance of excess credits that would go unused. The system learns from past credit usage patterns and adjusts accordingly.
3Adaptability or versatility
If the egress scheduler is located far from the virtual output queue, then the network architecture flexibility is improved, but the round-trip time increases causing credit waste
Solution Approach 1:
The patent changes the parameter used for threshold calculation from a fixed value to a dynamic parameter based on round-trip time. The threshold is recalculated as: threshold = RTT × credit_rate. This parameter change allows the system to accommodate varying distances between the virtual output queue and egress scheduler. Whether the scheduler is local or remote, the threshold automatically adjusts to the actual RTT, preventing credit waste while maintaining architectural flexibility.
Data Source
AI summary
A networking device controls transmission to an egress port with a virtual output queue using credits received from an egress scheduler. Credits may be requested at various rates based on the size of the virtual output queue. To optimize latency and prevent unused credits, the thresholds for changing the requested credit rates is determined based on the credit rate and the round-trip time from the virtual input queue to the egress scheduler. This ensures that when incoming data to the virtual output queue ends, the scheduler does not issue credits faster than the virtual output queue can signal changes to the effective credit rates.