Database Structure IO Pattern Prediction for Storage Tiering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing data storage systems face challenges in automatically tiering data based on database structures and operations, particularly for complex operations like drop/recreate, flip flop tables, and view switch, which complicate the recognition of associated activity and extent changes, leading to inefficient data placement.

Innovation Solution

A computer program that identifies defined database structures and operations, predicts associated activity, and provides hints to the storage array to proactively move data between tiers, utilizing historical data patterns and statistical analysis to anticipate and adjust data placement accordingly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional IO activity-based tiering is used, then data placement responds to actual access patterns, but complex database operations like drop/recreate, flip flop tables, and view switch cause delayed recognition and inefficient data placement

Engineering Contradiction:

Improvedata placement efficiencyVSAvoiddata movement delay

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by identifying database operations and predicting future IO activity before actual access patterns emerge. By analyzing database operation types (drop/recreate, flip flop tables, view switch) and predicting which extents will become hot, the system proactively moves data to appropriate tiers in advance, eliminating the delay inherent in traditional reactive tiering approaches.

Inventive Principle:

Principle #10Preliminary action

2Speed

If proactive data movement based on prediction is implemented, then data is moved to higher tiers before becoming active improving performance, but complexity in identifying and predicting database operation patterns increases

Engineering Contradiction:

Improvedata access speedVSAvoidprediction system complexity

Core Design Contradiction:

SpeedVSDevice complexity

Solution Approach 1:

The system segments the complex problem of predicting database IO patterns by breaking it down into distinct database operation types (drop/recreate, flip flop tables, view switch, etc.). Each operation type is analyzed independently with specific prediction rules, making the overall system more manageable and less complex than attempting to predict all IO patterns uniformly.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system changes parameters by monitoring specific database operation characteristics and using these parameter changes to trigger predictions. By tracking changes in database structure and operation types rather than attempting to predict all possible access patterns, the system reduces complexity while maintaining accurate predictions for known operation patterns.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS10162597B1Identifying IO access pattern in unique database structures

Publication Date: 2018.12.25 EMC IP HLDG CO LLC

AI summary

Certain database structures and operations are recognized in order to generate hints for a tiered storage array. Historical snapshots of a database data dictionary are searched for patterns corresponding to the structures and operations. After a structure or operation is identified, a corresponding set of extents is identified. Logic predicts future activity associated with the extents, and a corresponding hint is provided to the storage array. Statistical information associated with the identified structure or operation may be maintained and updated to facilitate prediction of activity.