OLTP Database Wait Time Estimation via Linear Regression

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In Online Transaction Processing (OLTP) databases, existing methods rely on fixed wait times for resource locking, leading to inefficiencies and inability to accurately predict when locks will be released, resulting in wasted time and potential process abortion.

Innovation Solution

A machine learning linear regression training platform determines database properties and trains a model using prior locking information to estimate wait times for OLTP database table access, enabling real-time predictions for wait times based on these properties.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If a fixed wait time is used for database locking, then the process can determine a predetermined time to wait, but this leads to wasted buffer time and inability to accurately predict lock release

Engineering Contradiction:

Improvewait timeVSAvoidwait time estimation accuracy

Core Design Contradiction:

Loss of timeVSMeasurement precision

Solution Approach 1:

The patent applies dynamics by transitioning from a static fixed wait time approach to a dynamic wait time estimation using machine learning. The system continuously learns from historical locking patterns and adjusts wait time predictions based on current database properties and locking information, making the wait time adaptive rather than predetermined.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent implements feedback by using historical locking information and actual wait time outcomes to train and refine the machine learning model. The system continuously incorporates new data about database access patterns, lock durations, and process behaviors to improve the accuracy of future wait time predictions.

Inventive Principle:

Principle #23Feedback

2Device complexity

If a fixed interval wait time is used, then the implementation is simple, but it causes processes to wait longer than necessary and waste computational resources

Engineering Contradiction:

Improvewait time mechanismVSAvoidprocess execution efficiency

Core Design Contradiction:

Device complexityVSProductivity

Solution Approach 1:

The patent applies preliminary action by training the machine learning model in advance using historical locking information and database properties. This pre-trained model can then quickly provide accurate wait time predictions without requiring complex real-time analysis, thus maintaining simplicity while improving efficiency.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent changes the parameter from a fixed constant wait time to a dynamic predicted wait time based on multiple variables including database properties, locking information, and learned patterns. This parameter transformation enables the system to adapt wait times to actual conditions rather than using a one-size-fits-all approach.

Inventive Principle:

Principle #35Parameter changes

3Ease of operation

If no wait time estimation is provided, then the system avoids complex prediction mechanisms, but processes cannot make informed decisions about whether to abort or wait

Engineering Contradiction:

Improvedecision-making capabilityVSAvoidprocess wait time

Core Design Contradiction:

Ease of operationVSLoss of time

Solution Approach 1:

The patent replaces the mechanical fixed-time waiting mechanism with an intelligent prediction system based on machine learning. This substitution enables the system to provide informed wait time estimates that help processes make better decisions about whether to wait or abort, replacing brute-force waiting with intelligent prediction.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

4Reliability

If the second process waits for the first process to finish, then resource locking is maintained, but buffer time is wasted when locks are released early

Engineering Contradiction:

Improveresource lockingVSAvoidwasted buffer time

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The patent uses feedback from historical locking patterns and actual lock release times to train the machine learning model. This feedback mechanism allows the system to learn when locks are typically released and adjust wait time predictions accordingly, reducing wasted buffer time while maintaining reliable resource locking.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11797511B2On-line transaction processing database wait time estimation using linear regression model

Publication Date: 2023.10.24 SAP SE

AI summary

Embodiments may be associated with database access. In some embodiments, a machine learning linear regression training platform determines a set of database properties (e.g., direct and/or indirect properties such as a document type, a type of process, a number of items, etc.) associated with Online Transaction Processing (“OLTP”) database table access. The machine learning linear regression training platform may then train a linear regression model based on the set of database properties and prior locking information (e.g., locking and unlocking timestamps) that represent access to the OLTP database table. Information about the linear regression model may be output to a wait time estimation platform that adapted to use the linear regression model to generate, in substantially real-time, an estimated wait time for an OLTP database table access based on the database properties of the OLTP database table access.