NVMe Submission Queue Priority Adjustment via Latency Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing NVMe protocol's arbitration mechanisms, such as round robin and weighted round robin, fail to effectively differentiate between critical and non-critical services, leading to potential delays and increased latency in submission queue execution, which can result in commands remaining unexecuted for long times.
Innovation Solution
A method is introduced where the host calculates latency periods and scores for submission queues, adjusting priorities dynamically to identify urgent queues and allocate higher priority to those with longer latency, thereby reducing the risk of commands being unexecuted by reconfiguring the arbitration mechanism to ensure timely processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If round robin arbitration mechanism is used, then all submission queues are treated equally, but critical services cannot be differentiated from non-critical services leading to delays
Solution Approach 1:
The patent implements dynamic priority adjustment by continuously monitoring latency scores of submission queues and reconfiguring arbitration weights in real-time. The host device dynamically calculates latency scores based on historical completion times and adjusts the priority weights of different submission queues accordingly, allowing the system to adapt to changing service requirements and differentiate between critical and non-critical services effectively.
Solution Approach 2:
The patent changes the arbitration parameter from fixed equal weights to dynamic weights based on latency scores. By introducing latency score as a new parameter and using it to modulate the arbitration weights of different submission queues, the system transforms the static round-robin mechanism into a dynamic priority-based arbitration system that can differentiate service levels.
2Reliability
If weighted round robin arbitration is used, then some queues get higher priority, but queues with large command volumes may receive incorrect priority adjustments
Solution Approach 1:
The patent implements a feedback mechanism where the host device continuously monitors the completion times of commands from different submission queues, calculates latency scores based on this feedback information, and uses the scores to adjust arbitration weights. This closed-loop feedback system ensures that priority adjustments are based on actual queue performance rather than static configurations, improving the accuracy of priority assignment while providing a systematic approach to handle large command volumes.
3Productivity
If static priority assignment is used, then arbitration mechanism is simple, but queues with longer latency cannot be identified and prioritized dynamically
Solution Approach 1:
The patent performs preliminary calculations of latency scores based on historical completion time data before making priority adjustments. By pre-calculating and storing latency scores for each submission queue, the system prepares the necessary information in advance for rapid priority reconfiguration, enabling quick response to latency issues without complex real-time calculations during arbitration decisions.
Data Source
AI summary
The present disclosure provides a method for adjusting a priority of a submission queue. The method includes: acquiring latency periods of a plurality of submission queues, wherein the latency periods of the submission queues are determined by an average value of completion periods of a plurality of commands in the submission queues, and the completion period is a difference between a timing at which an execution of the command is completed and a timing at which the command is added to the submission queue; determining latency scores of the submission queues based on the latency periods of the submission queues; determining a target submission queue among the plurality of submission queues based on the latency scores of the submission queues and adjusting the priority of the target submission queue; and sending an instruction to give notification of a result of a priority adjustment of the target submission queue.