Federated Machine Learning for Database Configuration Tuning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Database tuning in cloud platforms is challenging due to the lack of experienced administrators and limited historical optimization data for newly deployed databases, leading to suboptimal configuration settings that affect performance.

Innovation Solution

A computer-implemented method that selects database nodes with similar data factors in a centerless network, trains configuration models using data parallelism, combines them into a federated model, and recommends optimal configurations to new nodes based on performance thresholds, dynamically adjusting the network to improve performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If database configuration is manually tuned by administrators, then configuration quality can be improved, but it requires extensive administrative expertise that is often unavailable

Engineering Contradiction:

Improveconfiguration qualityVSAvoidadministrative expertise requirement

Core Design Contradiction:

Manufacturing precisionVSEase of operation

Solution Approach 1:

The database system performs automatic configuration tuning using machine learning models, eliminating the need for manual administrator intervention. The system self-diagnoses performance issues and self-adjusts configuration parameters based on learned patterns from historical data and federated knowledge from other database nodes.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical/manual process of administrator-based configuration tuning with an automated machine learning system. The ML model processes performance metrics and configuration parameters algorithmically, substituting human expertise with computational intelligence that can analyze data and make optimization decisions automatically.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Extent of automation

If auto-analysis feature is used with local optimization records, then configuration can be automatically filled, but historical optimization data is not abundantly available for new databases

Engineering Contradiction:

Improveautomatic configuration fillingVSAvoidhistorical optimization data availability

Core Design Contradiction:

Extent of automationVSLoss of information

Solution Approach 1:

The patent combines local optimization records with federated knowledge from multiple database nodes in a centerless network. By merging data from multiple sources through federated machine learning, the system creates a more comprehensive knowledge base that overcomes the limitation of insufficient local historical data, enabling effective automatic tuning even for newly deployed databases.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent transitions from single-database local optimization to multi-database federated optimization. By adding the dimension of cross-database knowledge sharing through a centerless network, the system accesses optimization patterns from diverse workloads and configurations, enriching the available optimization data beyond what any single database could provide locally.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

3Manufacturing precision

If database nodes are tuned individually, then local optimization can be achieved, but overall system performance and resource utilization are suboptimal

Engineering Contradiction:

Improvelocal optimization accuracyVSAvoidoverall system performance

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The patent creates a universal federated machine learning system that serves multiple database nodes simultaneously. The centerless network enables each node to benefit from collective knowledge while maintaining local customization, making the system multi-functional across diverse workloads and configurations rather than being limited to individual node optimization.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system implements continuous feedback loops where performance metrics from all database nodes are collected, analyzed, and used to update the federated machine learning models. This feedback mechanism enables the system to learn from actual runtime performance and continuously improve configuration recommendations, aligning local optimizations with overall system performance goals.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11138166B2Database tuning using a federated machine learning system of a centerless network

Publication Date: 2021.10.05 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Database configuration tuning is provided. A set of database nodes having similar data factors is selected in a centerless network of database nodes. Configuration models corresponding to the set of database nodes are trained using data parallelism. Trained configuration models corresponding to the set of database nodes are combined to form a federated configuration model. It is determined whether performance indicators corresponding to the set of database nodes are greater than a performance threshold level. In response to determining that the performance indicators corresponding to the set of database nodes are greater than the performance threshold level, a database configuration corresponding to the federated configuration model is recommended to a new database node. The new database node is joined to the centerless network.