HPC Authentication via Decentralized Primary Management Unit

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

High Performance Computing (HPC) machines face delays in computation due to reliance on centralized trusted third parties for security authentication, which compromises efficiency and introduces risks from malevolent hardware.

Innovation Solution

Implementing a decentralized authentication system using a primary management unit with a random data item generator and secure storage memory for each computing processor, allowing local generation and storage of a common secret data item for secure and encrypted data exchange, eliminating the need for a centralized trusted third party.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a centralized trusted third party is used for authentication, then security is maintained, but authentication delays increase and efficiency decreases

Engineering Contradiction:

ImprovesecurityVSAvoidauthentication delays

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The centralized authentication system is segmented into distributed authentication units located at each computing node. Each node possesses local authentication credentials (secret data items) that enable independent verification, eliminating the need for centralized authentication requests and reducing authentication delays while maintaining security through distributed verification

Inventive Principle:

Principle #1Segmentation

2Productivity

If hot maintenance is enabled for performance, then computing efficiency is improved, but security risks increase due to potential introduction of malevolent hardware

Engineering Contradiction:

Improvecomputing efficiencyVSAvoidsecurity risks from malevolent hardware

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

Authentication credentials are pre-distributed to each computing node before hot maintenance operations. When hardware is replaced during hot maintenance, the new component already possesses valid authentication credentials, enabling immediate security verification without interrupting system operation or requiring centralized authentication setup, thus maintaining both productivity and security

Inventive Principle:

Principle #10Preliminary action

3Reliability

If centralized authentication protocols are implemented, then security is ensured, but device complexity and energy consumption increase

Engineering Contradiction:

ImprovesecurityVSAvoidauthentication system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

Each computing node performs self-authentication using locally stored secret data items and public key infrastructure. Nodes independently verify each other's credentials without requiring complex centralized authentication protocols, reducing overall system complexity and energy consumption while maintaining security through decentralized self-service authentication

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS12124885B2High performance computing machine and method implemented in such a HPC machine

Publication Date: 2024.10.22 BULL SA

AI summary

A High Performance Computing (HPC) machine comprising several computing processors interconnected through at least one network, and at least one primary management unit, in a vicinity of at least one computing processor. The at least one primary management unit powers on the at least one processor. The at least one primary management unit (comprises a random data item generator, and a secure storage memory for storing a secret data item, common to all computing processors of the HPC machine, and used for authentication of each computing processor of the HPC machine for data exchange in the HPC machine.