NVMe/FC Submission Queue Bypassing Kernel Space
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large-scale distributed storage systems face performance issues due to latency associated with traversing the kernel stack when passing storage commands, which can impair business operations by increasing latency in data processing.
Innovation Solution
The system enables concurrent user space and kernel space access for storage commands through private, reserved I/O queues, allowing applications to bypass the kernel space for latency-sensitive operations like NVMe/FC commands, thereby reducing latency and improving data processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If storage commands traverse the kernel stack, then command processing can be centralized and controlled, but latency increases and data processing performance deteriorates
Solution Approach 1:
The patent segments I/O queue access into two distinct paths: traditional kernel space queues for controlled processing and private user space queues for low-latency access. This segmentation allows different command types to traverse different paths, resolving the contradiction between centralized control and latency sensitivity.
Solution Approach 2:
The patent introduces private I/O queues as an intermediary mechanism between user space applications and storage controllers. These private queues act as a mediator that eliminates the kernel stack traversal for latency-sensitive operations while maintaining the ability to process commands efficiently.
2Device complexity
If all storage commands go through kernel space, then centralized management is simplified, but data processing efficiency decreases due to increased latency
Solution Approach 1:
The patent applies local quality by providing different access mechanisms for different command types: private queues for latency-sensitive commands and traditional kernel queues for commands requiring centralized management. This localized differentiation optimizes both management complexity and processing efficiency without requiring complete system redesign.
3Loss of time
If private queues enable direct user space access, then latency is reduced, but queue management complexity increases
Solution Approach 1:
The patent implements multi-functionality by designing private I/O queues that can be shared across multiple user space applications while maintaining their low-latency characteristics. The queues serve multiple purposes: direct access path, concurrency control mechanism, and performance optimization tool, thereby managing complexity through functional consolidation.
Data Source
AI summary
A storage controller includes a processing device to send a Non-Volatile Memory Express over Fibre Channel (NVMe/FC) command to a submission queue without routing the NVMe/FC command through a kernel space, the submission queue being reserved for direct access by an initiator device to a user space of the storage controller.