NVMe Storage Mapping Table for Direct Write Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In distributed storage systems, the performance of NVMe storage devices is not fully utilized due to the need for central processing units (CPU) to process access requests, which hampers write performance and efficiency.
Innovation Solution
A management server is introduced to manage NVMe storage devices by maintaining a mapping table that maps storage node identifiers to memory addresses, allowing clients to directly access NVMe storage devices without involving the CPU in write requests, thereby improving write performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the CPU processes access requests to maintain system control and mapping, then system reliability is improved, but write performance deteriorates due to CPU involvement in every access operation
Solution Approach 1:
The patent extracts the CPU from the write path by pre-establishing mapping relationships between storage identifiers and queue start addresses. The CPU only processes read requests and mapping updates, while write requests bypass the CPU entirely and go directly to the NVMe storage device through the established mapping, thus improving write performance without compromising system reliability
Solution Approach 2:
The patent performs preliminary actions by pre-establishing and caching mapping relationships between storage identifiers and queue start addresses before write operations occur. This allows write requests to be processed directly using pre-computed mapping information, eliminating the need for CPU involvement during actual write operations
2Measurement precision
If the CPU parses access requests and converts addresses to maintain accurate mapping, then measurement precision is improved, but processing speed deteriorates due to CPU bottleneck
Solution Approach 1:
The patent performs address conversion and mapping establishment in advance during system initialization or when mapping changes occur. The pre-established mapping table caches the relationship between storage identifiers and queue start addresses, allowing subsequent write requests to use the cached mapping without CPU intervention, thus improving processing speed while maintaining accuracy
Solution Approach 2:
The patent creates a copy of the mapping relationship in the form of a cached mapping table that stores the correspondence between storage identifiers and queue start addresses. This copy allows direct address lookup without CPU-based conversion during write operations, improving speed while preserving the accuracy of the original mapping
3Ease of operation
If the client interacts with the storage node to obtain queue information, then system management is simplified, but access efficiency deteriorates due to additional interaction steps
Solution Approach 1:
The patent performs the action of obtaining queue information in advance by having the management server pre-fetch and cache the mapping relationships between storage node identifiers and queue start addresses. This preliminary action eliminates the need for clients to interact with storage nodes during write operations, improving access efficiency while maintaining simplified management through the caching mechanism
Data Source
AI summary
In a method for managing a storage device in a storage system, a client may send, based on an obtained start address that is of a queue of an NVMe storage device and to which an access request points and an obtained logical address that is of the NVMe storage device and to which the access request points, a remote direct memory access command to a storage node in which the NVMe storage device is located.