Ring Topology High-Performance Computing Network for Low Latency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing computer network systems are hindered by poor performance due to bandwidth and latency issues, particularly in inter-core communication, which affects the efficiency of parallel computations in finite element and finite volume methods.

Innovation Solution

A computer network system is designed with a high-performance computing network featuring a ring or daisy-chain topology for inter-thread communication, combined with a management network using a star topology for general communication needs. This setup includes high-speed network cards with infiniband communication protocol and DPU hardware, optimizing data processing and reducing latency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If communication occurs between cores distributed in different computers through network cards, then parallel computing capability is improved, but communication performance deteriorates due to bandwidth and latency limitations

Engineering Contradiction:
Improveparallel computing capabilityVSAvoidcommunication performance
Core Design Contradiction:
ProductivityVSSpeed

Solution Approach 1:

The patent segments the network into two distinct parts: a high-performance computing network for inter-thread communication and a management network for general communication. This segmentation allows each network to be optimized for its specific purpose, with the computing network using ring/daisy-chain topology and high-speed network cards dedicated to parallel computing tasks, thereby resolving the contradiction between parallel computing capability and communication performance.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces high-speed network cards with InfiniBand protocol as intermediaries between computational cores. These specialized network cards act as mediators that enable high-speed communication between distributed cores, overcoming the limitations of standard network cards and maintaining communication performance while enabling parallel computing across multiple computers.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of manufacture

If standard network cards are used for inter-computer communication, then system cost is reduced, but communication latency increases

Engineering Contradiction:
Improvesystem costVSAvoidcommunication latency
Core Design Contradiction:
Ease of manufactureVSLoss of time

Solution Approach 1:

The patent changes the key parameter of network card performance by specifying high-speed network cards with InfiniBand protocol for the computing network, while using standard network cards for the management network. This parameter differentiation allows the system to achieve low communication latency where needed (in inter-thread communication) while controlling costs in less critical areas (management communication).

Inventive Principle:
Principle #35Parameter changes

3Productivity

If computational domain is divided into many partitions for parallel processing, then computing speed is improved, but communication overhead increases due to data consistency requirements

Engineering Contradiction:
Improvecomputing speedVSAvoidcommunication overhead
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The patent applies local quality by creating a specialized high-performance communication environment specifically for inter-thread communication tasks. The ring/daisy-chain topology and high-speed network cards provide localized optimization for data exchange at partition interfaces, reducing communication overhead and energy loss while maintaining the benefits of fine-grained parallel processing across many partitions.

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS12284051B1Computer network system and computer system
Publication Date: 2025.04.22 ZHEJIANG UNIV
  • US12284051B1 patent drawing
  • US12284051B1 patent drawing
  • US12284051B1 patent drawing

AI summary

The present disclosure provides a computer network system. The computer network system includes: a high-performance computing network, where the high-performance computing network includes at least four computers, a topology relation of the large regions in the computational domain corresponds to a topology relation of the high-performance computing network, and the topology relation of the high-performance computing network is that the plurality of computers in the high-performance computing network form a ring topology or a daisy-chain topology by wired connection; where the management network includes all the computers in the high-performance computing network and computers having result post-processing for storage and computation, and all the computers in the management network are connected by switches or routers to form a star network topology. The present application provides a computer network system and a computer system, solves the problem of poor performance greatly influenced by bandwidth and latency.