NVMe over Fabric Packet Forwarding via Switch Path Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The integration of multipath software into operating systems for NVMe over Fabric architectures is complex due to software matching and compatibility issues across different operating systems, particularly in data centers where scalability and long-distance connections are needed, and existing solutions do not efficiently manage redundancy and path selection for reliable data transmission.

Innovation Solution

A forwarding and processing device that selects paths and encapsulates packets according to NVMe over Fabric requirements, allowing hosts to access storage resources using a generic NVMe protocol driver without the need for multipath software installation, thereby simplifying compatibility and reducing maintenance costs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If multipath software is integrated into the operating system for NVMe over Fabric, then path management and redundancy control are improved, but device complexity and software compatibility issues worsen

Engineering Contradiction:
Improvepath management reliabilityVSAvoidsoftware integration complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent introduces a switch as an intermediary device that handles path selection and multipath management functions. Instead of integrating complex multipath software into the host operating system, the switch acts as a mediator that receives NVMe commands from the host, selects appropriate paths to target storage devices, and manages redundancy at the network level. This transfers the complexity from the host software to the network infrastructure.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent extracts the multipath management functionality from the host operating system and relocates it to the switch. By taking out the path selection and redundancy control logic from the software layer and implementing it in the network device, the solution eliminates software compatibility issues while maintaining reliable path management.

Inventive Principle:
Principle #2Taking out (Extraction)

2Reliability

If multipath software is integrated into different operating systems, then path redundancy is improved, but ease of operation and maintenance worsen due to compatibility requirements

Engineering Contradiction:
Improveredundancy backupVSAvoidsoftware compatibility
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The switch serves as an intermediary that handles path management independently of the host operating system. This mediator approach allows the system to achieve redundancy backup without requiring operating system-specific multipath software, as the path selection logic resides in the network device rather than the host software.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The switch implements universal path management functionality that works across different operating systems. By providing a standardized interface and centralized control in the network device, the system achieves multi-functionality that supports various OS environments without requiring separate software implementations for each platform.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Speed

If PCIe protocol is used for NVMe connection, then transmission speed is improved, but scalability and long-distance connection capability worsen

Engineering Contradiction:
Improvedata transmission speedVSAvoidscalability and long-distance connection
Core Design Contradiction:
SpeedVSAdaptability or versatility

Solution Approach 1:

The patent merges the advantages of PCIe protocol for high-speed local connections with the scalability of network protocols. The host connects to the switch via PCIe for fast local access, while the switch provides network-based paths to target storage devices, combining the speed benefits of PCIe with the distance and scalability benefits of network infrastructure.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The solution transitions from a single-dimension PCIe direct connection to a multi-dimensional architecture that includes both PCIe and network paths. By adding the network dimension through the switch, the system maintains high-speed PCIe connectivity for local operations while gaining long-distance and scalable access through network protocols for remote storage devices.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP3757810B1Packet forwarding method, device, and system in nvme over fabric
Publication Date: 2023.04.05 HUAWEI TECH CO LTD
  • EP3757810B1 patent drawingFigure 1
  • EP3757810B1 patent drawingFigure 2
  • EP3757810B1 patent drawingFigure 3

AI summary

This application provides a packet forwarding method, device, and system in NVMe over Fabric, so as to simplify an implementation of multipath control in the NVMe over Fabric. According to the method provided in this application, a first packet sent by the control device is received, and an input/output command or a management command that complies with the NVMe protocol and that is included in the first packet is obtained; a second packet is generated according to the input/output command or the management command, where the input/output command or the management command complies with the NVMe protocol and is included in the first packet, and the second packet is a packet for executing the input/output command or the management command; a network port for forwarding the second packet is selected according to an identity of the target hard disk; and the second packet is encapsulated into a packet that meets an NVMe over Fabric forwarding requirement, and the encapsulated second packet is sent by using the selected network port. Complexity caused by a need of matching various possible operating systems during installation of multipath software on the control device is eliminated. In addition, maintenance costs of the control device are reduced.