Machine Learning Model Predicts Database Transaction Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Achieving low transaction latency in Database Management Systems (DBMS) is challenging due to various interacting factors such as resource bottlenecks, mutex locks, insufficient threads, storage, and network bandwidth issues, which become more complex as databases grow in size and complexity.
Innovation Solution
A method using machine learning models, specifically LASSO regression, to predict transaction latencies and optimize resource quota pool configuration settings by training models with performance records, allowing for user-adjustable settings and real-time recalculations based on current load and resource consumption values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If machine learning models are used to predict transaction latencies and optimize configurations, then transaction latency is reduced, but device complexity increases
Solution Approach 1:
The system performs preliminary actions by training machine learning models offline using historical performance records before actual query execution. The trained models predict transaction latencies and recommend configurations in advance, allowing the system to proactively optimize resource allocation and settings before latency issues manifest, thus reducing actual transaction latency without adding real-time computational complexity
Solution Approach 2:
The patent creates simplified copies of the complex database system by training machine learning models on historical performance data. These models serve as virtual representations that can predict latency and optimize configurations without requiring complex real-time analysis of the actual database system, thereby reducing transaction latency while managing device complexity through abstraction
2Productivity
If the database system grows in size and complexity to handle more data and concurrent users, then productivity increases, but transaction latency increases
Solution Approach 1:
The system implements feedback mechanisms by continuously monitoring database performance records and using machine learning models to analyze the relationship between system configuration, workload characteristics, and transaction latency. The models provide feedback recommendations for optimal configurations based on current system state and historical patterns, enabling the database to maintain low latency even as it scales to handle more concurrent users and larger data volumes
Solution Approach 2:
The patent applies dynamics by making the database configuration adaptive and changeable based on workload conditions. Machine learning models analyze performance records to dynamically recommend configuration settings that optimize transaction latency for different workload scenarios, allowing the system to adapt its behavior as it grows in size and complexity rather than remaining static
Data Source
AI summary
A processing system including at least one processor may obtain a first set of performance records of a database system, train a machine learning model in accordance with the first set of performance records, where the machine learning model that is trained in accordance with the first set of performance records is configured to predict a latency of a query transaction for a designated time period, present a user interface with a plurality of settings of the database system that are user-adjustable, where the plurality of settings is associated with at least a portion of the first set of performance records, calculate a first predicted latency of a query transaction at the designated time period via the machine learning model in accordance with a set of values of the plurality of settings, and present the first predicted latency via the user interface.