Network Interface Controller Message Striped Across Multiple Subnets

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current systems for high-performance computing face challenges in efficiently communicating large messages across multiple network interface controllers (NICs) while maintaining message ordering and resilience, particularly in large-scale systems with complex topologies and potential NIC failures, which can lead to issues like stale packets and increased costs due to infrastructure requirements.

Innovation Solution

A system that configures multiple NICs to communicate large messages using a manifest protocol, allowing for adaptive routing and fault tolerance by identifying alternate NICs in case of failures, and utilizing multiple subnets for increased bandwidth and resilience, ensuring message integrity and reducing infrastructure costs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multiple NICs are used to communicate large messages, then bandwidth and resilience are improved, but device complexity and infrastructure costs increase

Engineering Contradiction:
ImprovebandwidthVSAvoiddevice complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments large messages into smaller packets that can be transmitted across multiple NICs simultaneously. Each NIC handles a portion of the message traffic, dividing the communication load across multiple network interfaces. This segmentation enables parallel transmission paths, increasing overall bandwidth while managing complexity through structured packetization and routing protocols.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a manifest protocol that enables NICs to serve multiple functions: standard message transmission, failure detection, alternate path routing, and message reassembly. The same NIC infrastructure is used for both primary communication and fault tolerance mechanisms, reducing the need for dedicated redundant hardware while maintaining resilience.

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

2Reliability

If multiple NICs are deployed for fault tolerance, then reliability is improved, but infrastructure costs increase

Engineering Contradiction:
Improvefault toleranceVSAvoidinfrastructure costs
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent implements beforehand cushioning by pre-configuring manifest information that contains alternate NIC identifiers and routing paths before failures occur. When a NIC fails, the system can immediately switch to pre-identified alternate paths without requiring expensive dedicated backup hardware. The manifest protocol stores routing information that enables rapid failover, providing fault tolerance through software-based redundancy rather than hardware duplication.

Inventive Principle:
Principle #11Beforehand cushioning (Prior cushioning)

Solution Approach 2:

The patent uses copying by creating virtual representations of NIC identities and routing paths in the manifest protocol. Instead of requiring physical duplicate NICs for every possible failure scenario, the system copies routing information and NIC identifiers into the manifest, enabling logical redundancy that provides fault tolerance at minimal infrastructure cost.

Inventive Principle:
Principle #26Copying

3Reliability

If adaptive routing is implemented across multiple subnets, then message ordering and resilience are improved, but device complexity increases

Engineering Contradiction:
Improvemessage orderingVSAvoiddevice complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by embedding message sequencing information and routing instructions in the manifest protocol before transmission begins. The manifest contains pre-calculated routing paths and message order identifiers that guide packets through multiple subnets. This preliminary configuration enables adaptive routing decisions at network nodes without requiring complex real-time processing, maintaining message ordering while managing device complexity through pre-computed routing information.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS11799738B2Communication of a message using a network interface controller on a subnet
Publication Date: 2023.10.24 INTEL CORP
  • US11799738B2 patent drawing
  • US11799738B2 patent drawing
  • US11799738B2 patent drawing

AI summary

A system, method, and apparatus may provide one or more tangible, nontransitory computer-readable storage media having stored thereon executable instructions to instruct a processor to: stripe an outgoing network message into two or more pieces; send a first piece to a receiver via a first network interface card (NIC), and a second piece to the receiver via a second NIC; and upon determining that the receiver failed to receive a piece of the outgoing network message, replay the piece that the receiver failed to receive via a third NIC.