Cluster Node Configuration via Active System Identification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In high-performance computer clusters, configuring nodes in multiple boot mode often results in compatibility issues between operating systems and applications, leading to errors and significant latency during restarts due to incorrect configuration parameters.

Innovation Solution

Each node records an active system identifier in a database during boot, allowing for the automatic selection and application of compatible configuration parameters, ensuring seamless restarts by updating the cluster management database with the active system information and extracting relevant configuration parameters from the configuration management database.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If configuration parameters are manually set for each operating system, then compatibility between operating systems and applications is improved, but time consumption and latency during restarts increases significantly

Engineering Contradiction:
Improvecompatibility between operating systems and applicationsVSAvoidlatency during restarts
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system pre-configures multiple operating systems with their specific configuration parameters in the database before restart is needed. When a restart occurs, the management computer automatically detects the target operating system and applies the pre-prepared configuration parameters, eliminating manual configuration time and reducing restart latency while ensuring compatibility.

Inventive Principle:
Principle #10Preliminary action

2Speed

If configuration parameters are automatically selected without system identification, then restart speed is improved, but errors due to incorrect configuration parameters increase

Engineering Contradiction:
Improverestart speedVSAvoidaccuracy of configuration parameters
Core Design Contradiction:
SpeedVSReliability

Solution Approach 1:

The management computer implements a feedback mechanism where it detects the active operating system on each node, uses this information to automatically select the corresponding configuration parameters from the database, and applies them during restart. This closed-loop approach ensures both speed and accuracy by making configuration parameter selection dependent on actual system state.

Inventive Principle:
Principle #23Feedback

3Reliability

If extensive reboot preparation is performed to ensure correct settings, then compatibility and reliability are improved, but productivity and operational efficiency deteriorate

Engineering Contradiction:
Improvecorrectness of configuration settingsVSAvoidoperational efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system enables self-service automation where the management computer autonomously detects the operating system state, retrieves appropriate configuration parameters from the database, and applies them without requiring manual intervention. This automation maintains reliability while significantly improving productivity by eliminating time-consuming manual preparation steps.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP2851793B1Method for configuring at least one node of a computer cluster, corresponding equipment and corresponding system
Publication Date: 2018.12.19 BULL SA
  • EP2851793B1 patent drawingFigure 1
  • EP2851793B1 patent drawingFigure 2

AI summary

Method of configuring at least one node (200) of a cluster (1000) of computers, said at least one node having been deployed in multi-boot mode, characterized in that it comprises the following steps: - Booting (500) of the node, - Emission (501) by the node of a message (5010) identifying an active system, - Reception (503) by a management computer of the cluster of the identification message of an active system, - Updating (504) by the management computer of a database managing the computer cluster, of an identifier of an active system on the node, from the content of the identification message of an active system.