PCIe Interconnect via Network Switch for Low Latency Clustering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current high-speed interconnect schemes, such as Ethernet with TCP/IP protocols, suffer from high latency and packet drop issues under high traffic, limiting the growth of distributed computing and storage systems, and require protocol changes or data encapsulation, which introduce delays and costs.

Innovation Solution

The PCI Express (PCIe) standard is used for direct, high-speed data transfer between PCIe-enabled systems through PCIe-based peripherals connected via network switches, eliminating intermediate protocol conversions and allowing scalable bandwidth without architectural changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If Ethernet connection with TCP/IP protocols is used for high-speed data transfer, then transport speeds from 10 MB to 10 GB/sec are achieved, but high overhead and inherent latency make it unsuitable for some distributed applications

Engineering Contradiction:
Improvedata transfer speedVSAvoidlatency
Core Design Contradiction:
SpeedVSLoss of time

Solution Approach 1:

The patent extracts and removes the TCP/IP protocol layer from the data transfer path, allowing raw Ethernet frames to be transmitted directly between PCIe devices. This eliminates protocol overhead and reduces latency while maintaining high transfer speeds, directly resolving the contradiction between speed and latency.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces a PCIe bridge device as an intermediary that enables direct PCIe-to-Pcie communication over Ethernet infrastructure. This mediator translates PCIe transactions to Ethernet frames and back, allowing low-latency communication while utilizing existing Ethernet hardware, thus achieving both high speed and low latency.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of operation

If TCP/IP protocol is used for data transfer, then data can be transmitted between systems, but TCP/IP protocol tends to drop data packets under high traffic congestion which require resend and cause delays

Engineering Contradiction:
Improvedata transmission capabilityVSAvoidpacket delivery reliability
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The patent removes the TCP protocol layer that causes packet drops under congestion, using raw Ethernet frames instead. This eliminates TCP's flow control and retransmission mechanisms that lead to delays, providing more reliable and predictable data transfer for high-performance applications.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the communication parameter from TCP/IP protocol to raw Ethernet frames, fundamentally altering how data is transmitted. This parameter change eliminates protocol-induced packet drops and retransmissions, improving reliability and reducing latency under high traffic conditions.

Inventive Principle:
Principle #35Parameter changes

3Speed

If protocol changes or data encapsulation are made to achieve high-speed interconnect, then interconnect capability is improved, but associated delays increase and costs increase

Engineering Contradiction:
Improveinterconnect speedVSAvoidprocessing delay
Core Design Contradiction:
SpeedVSLoss of time

Solution Approach 1:

The patent extracts and eliminates unnecessary protocol conversion and data encapsulation steps in the interconnect path. By using raw Ethernet frames directly between PCIe devices, it removes the delays associated with protocol processing while maintaining high interconnect speeds.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The PCIe bridge device performs self-service by directly translating PCIe transactions to Ethernet frames without requiring external protocol conversion. This self-contained approach eliminates additional processing delays and reduces system complexity, achieving high speed with minimal latency.

Inventive Principle:
Principle #25Self-service

4Loss of time

If PCIe standard based peripheral are directly connected using PCIe standard based peripheral to PCIe standard based peripheral direct data link connections, then latencies of communication are reduced, but number of links per connection must be increased to meet bandwidth needs

Engineering Contradiction:
Improvecommunication latencyVSAvoidnumber of links
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent merges multiple PCIe links into a single logical connection through the PCIe bridge device. This consolidation maintains low latency by using direct PCIe-to-Pcie connections while reducing the apparent complexity by aggregating multiple physical links into one manageable interface for the host system.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The PCIe bridge device provides multi-functionality by handling both the physical link aggregation and the protocol translation functions. This universal device manages multiple links per connection internally while presenting a simplified interface externally, balancing low latency requirements with ease of configuration.

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

Data Source

PatentUS20230315674A1PCI Express to PCI Express based Low Latency Interconnect Scheme for Clustering Systems
Publication Date: 2023.10.05 THOMAS MAMMEN
  • US20230315674A1 patent drawing
  • US20230315674A1 patent drawing

AI summary

PCI Express (PCIE) is an IO interconnect standard that is implemented as a tree network with a root complex connecting to a CPU as the root node for use inside the computer for connecting to peripheral devices. Currently PCIE has achieved stability such that PCIE can be used as a basis for other applications. A PCIE based scheme inter-connecting multiple PCIE enabled computer systems each having at least one PCIE root complex controlling at least a PCIE bus, enabling the scalability of PCIE architecture to be applied for data transport between the connected system cluster is proposed. The interconnection uses an outbound port enabled for system interconnection on the PCIE bus of each PCIE enabled computer connecting to one inbound port on an independently programmable network switch having a plurality of inbound ports. The interconnection is using PCIE protocol for data transfer within the cluster.