Single-Module Table Designation for Parallel Database Scalability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Parallel database systems face scalability issues with small tables and single row operations due to skewed distribution of rows across access modules, leading to high coordination overheads and inefficient processing.
Innovation Solution
A database system that allows designation of tables as single-module tables, where rows are collectively assigned to a single access module, reducing coordination overheads and improving processing efficiency by using a processor to dynamically select the owning access module based on processing costs and load balancing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If rows are distributed across multiple access modules in parallel, then processing speed is improved for large tables, but coordination overhead increases for small tables and single row operations
Solution Approach 1:
The system dynamically determines whether to use single-module or multi-module table processing based on the specific operation type and table size. The processor selectively designates tables as single-module tables for small tables and single row operations, while maintaining multi-module distribution for large tables, allowing the system to adapt its processing mode to optimize performance for different workloads
Solution Approach 2:
The patent applies different processing strategies to different types of tables and operations. Small tables and single-row operation tables are designated as single-module tables with concentrated row access, while large tables maintain distributed multi-module access. This local differentiation allows each table type to be processed in the most efficient manner for its specific characteristics
2Productivity
If rows of a small table are distributed across multiple access modules, then parallel access is enabled, but row distribution becomes skewed and processing efficiency decreases
Solution Approach 1:
For small tables and tables requiring single row operations, the patent merges all rows into a single access module rather than distributing them across multiple modules. This consolidation eliminates the skew problem inherent in distributing small datasets across many modules and allows the single module to process all rows efficiently without coordination overhead
3Adaptability or versatility
If single row operations are processed in a shared nothing MPP system with multiple access modules, then system scalability is maintained, but coordination overhead and messaging between modules increases
Solution Approach 1:
The system dynamically selects the appropriate processing mode based on operation characteristics. For single row operations, it designates the table as a single-module table, allowing one access module to handle the entire operation without inter-module messaging. For operations requiring system scalability, the multi-module architecture remains available, creating a flexible system that adapts to different operational requirements
Data Source
AI summary
There is provided a parallel database system 1. System 1 includes a storage device 2 for maintaining a plurality of rows 3. System 1 also includes a plurality of access modules 4 for accessing in parallel a respectively assigned one or more of rows 3. An interface 5 receives data 6 indicative of a table 7 defined by a selection 8 of rows 3. A processor 9 is responsive to data 6 for selectively designating table 7 to be a single-module table. In this single module table, selection 8 is collectively assigned to an owning one of modules 4, indicated by reference numeral 10.