Bridge Buffer Transfer for Host-Free NVMe-oF Storage Arrays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data transfer technologies involving Non-Volatile Memory Express over Fabrics (NVMe-oF) face inefficiencies when transferring data between storage devices connected to the same bridge device, as they require data to be routed through the host system, creating bottlenecks in high-throughput scenarios.
Innovation Solution
A bridge device that converts fused read/write data transfer commands into separate read and write commands, utilizing a buffer memory not located in the host system, enabling direct data transfer between storage devices without involving the host system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data is transferred through the host system, then compatibility with existing NVMe-oF protocols is maintained, but data transfer efficiency and throughput are reduced
Solution Approach 1:
The bridge device introduces an intermediary buffer memory that enables direct data transfer between source and destination storage devices without routing through the host system. This mediator approach resolves the contradiction by providing a dedicated data path that maintains protocol compatibility while eliminating the performance bottleneck of host-mediated transfers.
Solution Approach 2:
The data transfer operation is segmented into distinct phases: command transmission to the bridge device, buffer allocation, data read from source device to buffer, and data write from buffer to destination device. This segmentation allows the bridge device to manage the transfer process independently, improving efficiency while maintaining controlled complexity through modular operation.
2Reliability
If data is transferred via host system, then protocol compatibility is ensured, but transfer time and latency increase
Solution Approach 1:
The bridge device serves as an intermediary that understands NVMe-oF protocols from the host and translates them into direct storage device commands. This mediator function preserves protocol compatibility for host communication while enabling time-efficient direct data paths between storage devices, eliminating the need for data to traverse the host memory and CPU.
3Adaptability or versatility
If host system is involved in data transfer, then command processing capability is maintained, but system resource consumption increases
Solution Approach 1:
The bridge device extracts the data transfer processing function from the host system and relocates it to the storage subsystem. By taking out the heavy lifting of data movement from the host, the solution maintains command processing capability through the bridge device's controller while significantly reducing host resource consumption for data transfer operations.
Data Source
AI summary
Systems, methods, and a bridge device for a data storage device array that enables data transfer between data storage devices without transferring data to the host system are described. The bridge device receives a data transfer command from the host system, such as a fused read and write command, and converts it to separate read and write commands for the source and destination storage devices. The bridge device may operate in a non-volatile memory express over fabric (NVMe-oF) environment and manage the command queues, connection verifications, and command conversions to enable data transfer among storage devices connected to a shared peripheral component interface express (PCIe) root complex without using host network bandwidth or memory.