Table Distribution Optimization for Distributed Database Systems

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Distributed database systems face challenges in optimizing table distributions among nodes to improve performance and resource utilization, particularly in reducing performance overhead and achieving even computing resource use during database operations involving multiple nodes.

Innovation Solution

A method is provided to determine optimal table distributions by iteratively analyzing candidate distributions using a scoring function that incorporates performance metrics, such as the number of two-phase commit operations and computing resource distribution, until a stopping condition is met, and automatically redistributing tables among nodes to achieve the final distribution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If tables are distributed across multiple nodes in a distributed database system, then data storage capacity and system scalability are improved, but performance overhead increases due to multiple two-phase commit operations and multinode select operations

Engineering Contradiction:

Improvedata storage capacityVSAvoidperformance overhead

Core Design Contradiction:

Quantity of substanceVSLoss of energy

Solution Approach 1:

The system changes the parameter of table distribution configuration by analyzing execution data and using optimization algorithms (such as simulated annealing) to determine optimal distribution patterns that minimize two-phase commit operations while maintaining data storage capacity across multiple nodes

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If tables are distributed across multiple nodes, then system scalability is improved, but computing resource use becomes uneven across nodes

Engineering Contradiction:

Improvesystem scalabilityVSAvoidevenness of computing resource use

Core Design Contradiction:

Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The system implements feedback mechanisms by continuously analyzing database execution data to monitor computing resource usage patterns across nodes, then uses this feedback to iteratively optimize table distributions and achieve more even resource utilization while preserving system scalability

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The table distribution is made dynamic through iterative optimization processes that adjust distributions based on analyzed execution data, allowing the system to adaptively balance computing resources across nodes while maintaining scalability

Inventive Principle:

Principle #15Dynamics

3Ease of operation

If manual table distribution configuration is used, then implementation simplicity is maintained, but performance optimization capability is limited

Engineering Contradiction:

Improveimplementation simplicityVSAvoidperformance optimization capability

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The system performs self-service by automatically analyzing its own execution data and using optimization algorithms to determine optimal table distributions, eliminating the need for manual configuration while achieving significant performance improvements through reduced two-phase commit operations

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS11429584B2Automatic determination of table distribution for multinode, distributed database systems

Publication Date: 2022.08.30 SAP SE

AI summary

Technologies are described for improving the performance of distributed database systems. The technologies provide suggested table distributions that can reduce the performance overhead associated with database operations that involve multiple nodes, and can provide more even distribution of computing resource use. In a particular example, a method includes minimizing a scoring function that includes a plurality of performance metrics, at least a portion of the metrics being associated with a weight. Candidate table distributions can be analyzed in an iterative manner to determine a lowest-score distribution. A final table distribution can be output, such as to a user.