Cluster Component Inventory Automation via Distributed Agents

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The growing complexity of cluster components and their varying sizes require more efficient means of administration and management, particularly in high-performance computing environments where manual inventory processes are inefficient and prone to errors.

Innovation Solution

A method and computer program for dynamically identifying and managing cluster components by performing automatic local and global inventories, using local inventory modules on nodes with data processing capabilities and a summary inventory module on management nodes to aggregate and update component information, enabling optimized cluster management operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual inventory processes are used to manage cluster components, then operators can query and document component information, but the process becomes inefficient and prone to errors as cluster complexity increases

Engineering Contradiction:
Improveinventory accuracyVSAvoidinventory efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The cluster components perform self-inventory through automated agents running on each node. These agents automatically detect and report hardware and software components without requiring manual operator intervention, thereby improving both reliability (no human error) and productivity (automated continuous monitoring) of the inventory process

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The manual mechanical process of operators physically querying and documenting components is replaced by an automated electronic system using software agents that communicate with cluster components through standardized interfaces, eliminating human error and significantly improving inventory efficiency

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Quantity of substance

If the cluster size increases from dozens to thousands of nodes, then computing power increases, but the complexity of component inventory and management increases significantly

Engineering Contradiction:
Improvenumber of nodesVSAvoidmanagement complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The inventory management system is segmented into distributed autonomous agents on each node and a centralized management server. Each node's agent independently manages its local component inventory, reducing the complexity burden on the central system and enabling scalable management of large clusters with thousands of nodes

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

A universal inventory agent is designed to work across all node types (compute, storage, administration nodes) regardless of their specific functions. This multi-functional agent handles diverse component types uniformly, simplifying management complexity as the cluster scales

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

3Extent of automation

If remote automatic inventory is implemented, then inventory can be carried out automatically in many configurations, but new methods for automating management operations are required

Engineering Contradiction:
Improveinventory automationVSAvoidsystem complexity
Core Design Contradiction:
Extent of automationVSDevice complexity

Solution Approach 1:

An intermediary inventory agent is introduced between the remote management system and cluster components. This agent standardizes communication protocols and data formats, enabling automatic inventory across diverse configurations while managing system complexity through a unified interface layer

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP2727057B1Method and computer program for dynamically identifying components of a cluster and for automating operations for optimized management of the cluster
Publication Date: 2019.04.24 BULL SA
  • EP2727057B1 patent drawingFigure 1
  • EP2727057B1 patent drawingFigure 2~3
  • EP2727057B1 patent drawingFigure 4~5

AI summary

The invention relates in particular to the dynamic identification of components of elements (305, 310) of a cluster (300) so as to automate operations for optimized management of the cluster. After having been established in a first element, a local inventory comprising characteristics of hardware and/or software components of the first element is transmitted to a second element, distinct from the first. A partial or global inventory comprising characteristics of hardware and/or software components of the first element is then established in the second element and an operation of managing the cluster is launched according to said partial or global inventory.