Dual-Tree PCIe Backplane for NVMe Traffic Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing backplanes for NVMe drives in information handling systems face challenges with scalability and redundancy, as they are based on point-to-point PCIe topology, leading to unbalanced storage traffic and lack of failover redundancy.
Innovation Solution
A dual-tree topology fabric is established using specially designed expandable PCIe backplanes connected in a chain configuration, allowing storage traffic to be balanced between multiple uplink paths and providing redundancy links for each NVMe drive through dual-port connections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a point-to-point PCIe topology is used for NVMe drives, then the connection structure is simple, but storage traffic cannot be balanced and no redundancy links are available
Solution Approach 1:
The system is divided into multiple PCIe root complexes (at least two) instead of using a single root complex. Each NVMe drive is segmented to have separate connections to different root complexes, creating multiple independent paths. This segmentation enables both traffic balancing across multiple paths and redundancy if one path fails.
Solution Approach 2:
The topology transitions from a single-dimension point-to-point connection to a multi-dimensional mesh-like structure where drives can be reached through multiple dimensional paths (different root complexes). This adds spatial redundancy to the connection architecture, allowing traffic to route through alternative paths when needed.
2Quantity of substance
If expandable backplanes are used to connect a large number of drives, then the system capacity increases, but scalability and extensibility become serious issues
Solution Approach 1:
The backplane system is segmented into multiple modular backplane units, each connected to different PCIe root complexes. This allows the system to scale by adding more backplane modules without overloading a single PCIe switch, improving scalability while maintaining the ability to connect a large number of drives.
Solution Approach 2:
The backplane design incorporates universal PCIe interfaces that can accommodate different types of storage drives and configurations. The multi-root complex architecture provides multi-functionality by supporting both high-capacity configurations and scalable expansions, making the system adaptable to various deployment scenarios.
3Productivity
If NVMe drives are placed in leaf nodes of a PCIe-based tree topology, then the protocol advantages are utilized, but storage traffic cannot be balanced between multiple cable paths
Solution Approach 1:
The PCIe topology is segmented into multiple root complex trees instead of a single tree. NVMe drives positioned at leaf nodes can be assigned to different root complexes, enabling traffic balancing across multiple independent PCIe paths. This maintains NVMe protocol performance while adding load distribution capability.
Solution Approach 2:
The system dynamically routes NVMe traffic through different root complexes based on load conditions. The multi-path architecture allows flexible traffic distribution, enabling the system to adaptively balance storage traffic across available paths while maintaining optimal NVMe protocol performance.
Data Source
AI summary
An information handling system may include at least one processor; a first and a second backplane, wherein the first and second backplanes are Peripheral Component Interconnect Express (PCIe) backplanes; and a physical storage resource. The physical storage resource may be coupled to the at least one processor via a first port of the physical storage resource and via the first backplane, and the physical storage resource may be further coupled to the at least one processor via a second port of the physical storage resource and via the second backplane.