Graph Data Task Scheduling via Distributed Quantized Value Exchange

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional centralized task scheduling methods in graph-data-based computing face inefficiencies and reliability issues, especially with large data volumes, leading to low processing efficiency and potential errors due to reliance on a single parameter server.

Innovation Solution

A graph-data-based task scheduling method that divides to-be-processed graph data into subgraph data, using quantized values to measure computing capacity and load status, allowing graph nodes to dynamically exchange resources and adjust allocations to maintain equilibrium, thereby implementing self-organized load balancing and avoiding single point failures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If a centralized parameter server is used for unified scheduling and control of computing tasks, then task management is simplified and centralized, but processing efficiency deteriorates when data volume is large and reliability is compromised

Engineering Contradiction:
Improvetask management complexityVSAvoidgraph data processing efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent divides the centralized parameter server into multiple distributed computing nodes, each capable of independent task execution and scheduling. The graph data is segmented into subgraph data distributed across multiple nodes, eliminating the single point of failure and bottleneck in centralized architectures.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a blockchain-based quantized value system as an intermediary mechanism for resource allocation and task scheduling. This mediator enables automated, trustless coordination between distributed nodes without requiring a centralized controller, resolving the contradiction between centralization simplicity and distributed scalability.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of operation

If a centralized parameter server is used for unified scheduling, then task coordination is centralized, but reliability deteriorates due to single point of failure

Engineering Contradiction:
Improvetask coordinationVSAvoidfault tolerance
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The system segments the centralized coordination function into multiple distributed nodes that each maintain local task coordination capabilities. This segmentation eliminates the single point of failure while preserving coordination functionality through peer-to-peer interactions.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transforms the coordination mechanism from centralized control to distributed consensus based on quantized values recorded on blockchain. This parameter change in the coordination approach maintains operational simplicity while dramatically improving reliability through decentralization.

Inventive Principle:
Principle #35Parameter changes

3Quantity of substance

If graph data is processed in large volumes through centralized scheduling, then data processing capacity is consolidated, but scheduling efficiency deteriorates

Engineering Contradiction:
Improvedata processing capacityVSAvoidscheduling efficiency
Core Design Contradiction:
Quantity of substanceVSProductivity

Solution Approach 1:

The patent segments large-scale graph data into smaller subgraph units that can be independently processed by multiple computing nodes. This segmentation enables parallel processing while maintaining scheduling efficiency through distributed resource allocation based on quantized values.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system implements dynamic task scheduling where computing nodes can dynamically acquire and release graph nodes based on their current load and capabilities, represented through quantized values. This dynamic approach maintains high scheduling efficiency even when processing large volumes of data.

Inventive Principle:
Principle #15Dynamics

4Productivity

If computing nodes dynamically exchange graph nodes and quantized values, then load balancing is improved, but system complexity increases

Engineering Contradiction:
Improvetask scheduling efficiencyVSAvoidresource allocation mechanism
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements a feedback mechanism where computing nodes continuously monitor their load status through quantized values and automatically adjust resource allocation accordingly. This feedback-driven approach achieves load balancing through simple local decisions rather than complex centralized control.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

Computing nodes autonomously perform load balancing by exchanging graph nodes and quantized values with other nodes based on their own resource status. This self-service mechanism eliminates the need for complex external coordination while achieving efficient load distribution.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3796166B1Graph data-based task scheduling method, device, storage medium and apparatus
Publication Date: 2024.03.06 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP3796166B1 patent drawingFigure 1
  • EP3796166B1 patent drawingFigure 2~3
  • EP3796166B1 patent drawingFigure 4~6

AI summary

This application relates to a graph data-based task scheduling method, device, storage medium and apparatus. The method comprises: acquiring subgraph data obtained by dividing graph data to be processed; performing a unit computing task on graph nodes in the subgraph data; acquiring a first quantization value, which is a quantization value corresponding to a completed unit computing task when data corresponding to the completed unit computing task is formed into blockchain data; acquiring a second quantization value, which is a quantization value corresponding to the completed unit computing task when the data corresponding to the completed unit computing task is not formed into the blockchain data; determining a third quantization value corresponding to an uncompleted unit computing task in the subgraph data; and exchanging the graph nodes and the first quantization value with other computing nodes when the second quantization value and the third quantization value do not meet an equalization condition. The solution provided in this application can improve the efficiency of task scheduling.