Forecasting Evolving Database Query Workloads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current database management systems (DBMS) face challenges in accurately capturing and forecasting evolving query workloads due to their static assumptions, leading to ineffective recommendations for future workloads, as historical data fails to account for time-evolving patterns and changing predicate parameters.
Innovation Solution
A system and method for characterizing and forecasting evolving query workloads using a template-based featurization approach with a deep learning model, which identifies recurrent queries, generates feature vectors, and predicts future queries and their arrival times, employing an encoder-decoder architecture and a cut and pack technique to optimize model capacity and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If historical query traces are used as input workload for physical design tools, then the tools can automatically recommend physical design features like indexes and materialized views, but the recommendations are not always effective for future workloads because real workloads are dynamic and exhibit time-evolving behavior
Solution Approach 1:
The patent applies dynamics by transforming the static historical workload into a dynamic forecasted workload that evolves over time. The system uses time-series forecasting models to generate future query workloads that capture temporal patterns and changes, allowing physical design recommendations to adapt to evolving workload characteristics rather than relying on static historical data
Solution Approach 2:
The patent applies preliminary action by forecasting future query workloads before they actually occur. The system predicts upcoming query patterns, parameter values, and workload characteristics in advance, enabling physical design tools to prepare optimized recommendations proactively rather than reactively after workload changes have already happened
2Productivity
If recommended views are generated based on queries with old predicate parameter values, then the views can be created efficiently, but the views may not be useful for future queries if the predicate parameters change over time
Solution Approach 1:
The patent applies dynamics by generating view recommendations based on forecasted future parameter values rather than static historical values. The system predicts how predicate parameters will evolve and creates views that are optimized for future query patterns, ensuring views remain useful as workloads change over time
Solution Approach 2:
The patent applies parameter changes by explicitly modeling and predicting changes in query parameter values over time. The system uses forecasting models to predict future parameter values and incorporates these predictions into view recommendation generation, allowing views to be optimized for evolving parameter distributions rather than fixed historical values
3Measurement precision
If deep learning models are used to forecast query workloads, then the forecasting accuracy improves, but the model capacity and computational requirements increase
Solution Approach 1:
The patent applies segmentation by decomposing complex query workload forecasting into multiple independent template-level forecasting tasks. Instead of forecasting entire queries as a single complex sequence, the system segments queries into templates and parameters, forecasting each template's parameter values separately. This reduces model complexity while maintaining overall forecasting accuracy
Solution Approach 2:
The patent applies universality by designing forecasting models that can handle multiple query templates and parameter types with a unified approach. The system uses generalizable model architectures that work across different database schemas and query patterns, reducing the need for separate specialized models for each template type and thereby reducing overall system complexity
Data Source
AI summary
Systems and methods for characterizing and forecasting evolving query workloads. The method includes receiving a query, the received query including a parameter value and an arrival time; identifying the query as a recurrent query; extracting a query template from the received query by parsing the received query; based at least on the identifying, generating a feature vector for the received query, the feature vector generated based on the extracted template and the parameter value; and forecasting a future query based on the generated feature vector by applying a neural network.