Shared-Nothing Data Store Expansion via Scheduled Redistribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Expanding a shared-nothing data store system requires significant downtime due to the time-consuming process of redistributing data across nodes, which can take several days and disrupts service.
Innovation Solution
A data redistribution schedule table with specified parameters is created to manage the redistribution of data across nodes, allowing for minimal downtime by prioritizing frequently used tables and performing redistribution during off-peak hours, while maintaining data integrity and accessibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If data redistribution is performed across all nodes when expanding a shared-nothing system, then the system capacity and scalability are improved, but the service downtime increases significantly
Solution Approach 1:
The patent segments the data redistribution process by identifying and prioritizing frequently accessed tables for redistribution to new nodes, while leaving less frequently accessed tables on existing nodes. This selective segmentation allows the system to expand capacity without requiring complete data redistribution across all nodes, thereby reducing service downtime while maintaining scalability.
Solution Approach 2:
The patent performs preliminary identification of frequently accessed tables and pre-plans their redistribution to new nodes before the expansion operation. By preparing the redistribution schedule and identifying target tables in advance, the system minimizes the actual downtime during the expansion event while still achieving the desired system capacity increase.
2Manufacturing precision
If complete data redistribution is performed across the system, then data distribution uniformity is improved, but the time required for redistribution increases to several days
Solution Approach 1:
The patent applies partial action by redistributing only the subset of frequently accessed tables that require uniform distribution, rather than redistributing all tables in the system. This selective approach achieves sufficient data distribution uniformity for critical operations while dramatically reducing the total redistribution time from several days to a much shorter duration.
Solution Approach 2:
The patent applies different redistribution strategies to different types of tables based on their access patterns. Frequently accessed tables receive priority redistribution to ensure uniformity and performance, while less frequently accessed tables maintain their existing distribution. This local quality differentiation optimizes the balance between distribution uniformity and redistribution time.
3Productivity
If data redistribution is performed during peak hours, then system maintenance can be completed faster, but user access and database operations are disrupted
Solution Approach 1:
The patent performs preliminary identification and prioritization of frequently accessed tables before the redistribution operation. By preparing the redistribution schedule in advance and identifying which tables require redistribution, the system can execute the actual data movement during off-peak hours with minimal disruption, while still achieving the productivity goal of completing maintenance during scheduled windows.
Solution Approach 2:
The patent implements periodic or scheduled redistribution operations during off-peak hours rather than attempting to complete all redistribution during a single peak-hour window. This periodic approach allows data redistribution to proceed when user access is lower, maintaining service availability while still achieving the necessary redistribution speed and productivity targets over time.
Data Source
AI summary
A computer readable storage medium includes executable instructions to evaluate an expanded shared-nothing data store configuration. A data redistribution schedule table with specified parameters is formed. Data is redistributed within the expanded shared-nothing data store in accordance with the data redistribution schedule table.