Supercomputer Network Subnetwork Allocation for Link Congestion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for deploying tasks in supercomputers do not effectively control the task execution environment, leading to congestion and inefficiencies in network usage, as well as high energy consumption due to unnecessary switch activation.

Innovation Solution

A method that allocates nodes and links in a supercomputer network to form a sub-network that uses only unallocated or minimally allocated links, reducing congestion and energy consumption by implementing inter-node communication routes within the allocated sub-network and placing unused links in an inactive state.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If multiple tasks share the same network links in a supercomputer, then network resource utilization is improved, but network congestion and task execution variability increase

Engineering Contradiction:
Improvenetwork resource utilizationVSAvoidtask execution environment stability
Core Design Contradiction:
Use of energy by moving objectVSReliability

Solution Approach 1:

The patent segments the network into multiple virtual networks (subnetworks) that are allocated to different tasks. Each task receives a dedicated virtual network with guaranteed bandwidth, preventing network congestion while maintaining high resource utilization through efficient virtualization. This resolves the contradiction by dividing the shared physical network into isolated logical segments.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a network controller as an intermediary that manages and allocates network resources between multiple tasks. The controller dynamically assigns virtual network resources to tasks, ensuring fair distribution and preventing any single task from monopolizing network bandwidth, thus maintaining both high utilization and stable execution environments.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of operation

If the supercomputer maintains all network links active to ensure task execution, then task availability is improved, but energy consumption increases

Engineering Contradiction:
Improvetask execution availabilityVSAvoidnetwork link energy consumption
Core Design Contradiction:
Ease of operationVSUse of energy by stationary object

Solution Approach 1:

The patent implements dynamic network link management where the state of network links (active or inactive) changes based on task requirements. When tasks are allocated to specific nodes, only the necessary network links are activated, and unused links are deactivated to save energy. This dynamic adaptation resolves the contradiction between availability and energy consumption.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The network system automatically manages link activation and deactivation based on task allocation patterns. The system monitors task execution requirements and self-adjusts network resource states without manual intervention, ensuring tasks have available network resources while minimizing energy consumption by keeping only necessary links active.

Inventive Principle:
Principle #25Self-service

3Productivity

If network routing is dynamically recalculated to avoid congestion, then network efficiency is improved, but deployment time and system complexity increase

Engineering Contradiction:
Improvenetwork communication efficiencyVSAvoidrouting management complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent pre-calculates and establishes virtual network routes during task allocation before actual data transmission begins. By determining the network path in advance and reserving resources along that path, the system avoids the need for complex real-time routing recalculations during task execution, thus maintaining high efficiency while reducing operational complexity.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3671455A1Method for deploying a task in a supercomputer, method for implementing a task in a supercomputer, corresponding computer program and supercomputer
Publication Date: 2020.06.24 BULL SA
  • EP3671455A1 patent drawingFigure 1
  • EP3671455A1 patent drawingFigure 2
  • EP3671455A1 patent drawingFigure 3

AI summary

This task deployment process involves: allocating nodes (1...16) to the task; determining, within the network (110), a subnetwork for interconnecting the allocated nodes, satisfying one or more predefined criteria, including a first criterion whereby the subnetwork uses only links that are not allocated to any other task already deployed, or that are allocated to fewer than N other tasks already deployed, where N is a predefined number of one or more; allocating the subnetwork, and in particular the links belonging to this subnetwork, to the task; and implementing inter-node communication routes within the allocated subnetwork.