Multi-Host RDMA Card CPU Coupling via Midplane

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional approaches to communication between storage processors, such as using PCIe non-transparent bridges, lack compliance across CPU vendors and generations, fail to support virtualized systems, and require high software overhead to achieve reliable messaging, while also necessitating physical cable connections and blocking IO slots.

Innovation Solution

A multi-host RDMA card is coupled to a pair of CPUs, providing Ethernet over PCIe connectivity and mirroring data between them, eliminating the need for physical cables and supporting virtualized systems without software stacks, and allowing high-speed communication across storage processors.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If PCIe non-transparent bridge is used for communication between storage processors, then physical connection is established, but compatibility across different CPU vendors and generations is poor

Engineering Contradiction:
Improvecommunication reliabilityVSAvoidCPU vendor and generation compatibility
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent introduces a midplane as an intermediary component between CPUs and IO cards. The midplane provides a standardized interface that mediates communication between different CPU vendors and generations, eliminating direct CPU-to-CPU dependencies and enabling universal compatibility while maintaining reliable physical connections.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The midplane is designed with universal connectivity capabilities that support multiple CPU vendors and generations through standardized interfaces. This multi-functional design allows the same midplane structure to accommodate different CPU types without requiring vendor-specific or generation-specific modifications.

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

2Reliability

If conventional IO card approach with external cables is used, then dedicated communication path is provided, but device complexity and space requirements increase

Engineering Contradiction:
Improvecommunication reliabilityVSAvoidnumber of IO cards and physical cables
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the communication functionality into the midplane structure itself, eliminating the need for separate external cables and multiple IO cards. The midplane integrates both the connection infrastructure and communication pathways, reducing overall device complexity while maintaining dedicated communication paths between storage processors.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent extracts the communication pathway from external physical cables and integrates it directly into the midplane structure. This extraction eliminates the need for external cable management and reduces the number of physical components required while maintaining reliable dedicated communication paths.

Inventive Principle:
Principle #2Taking out (Extraction)

3Adaptability or versatility

If conventional approaches are used for virtualized systems, then support for virtual machines is required, but software overhead increases

Engineering Contradiction:
Improvevirtualized system supportVSAvoidsoftware stack overhead
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The midplane provides self-service capabilities by implementing built-in support for virtualized systems and virtual machines. Rather than requiring external software stacks to enable virtualization support, the midplane inherently provides the necessary functionality, eliminating software overhead while maintaining adaptability to virtualized environments.

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS11675719B2System and method for multi-node communication
Publication Date: 2023.06.13 EMC IP HLDG CO LLC
  • US11675719B2 patent drawing
  • US11675719B2 patent drawing
  • US11675719B2 patent drawing

AI summary

A method, computer program product, and computing system for coupling a multi-host remote direct memory access (RDMA) card to at least a pair of central processing units (CPUs). One or more signals may be routed, via the multi-host RDMA card, between the at least a pair of CPUs.