Non-homogenous Synopsis for Partition Pruning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing in-memory databases face inefficiencies in query processing due to uniform synopsis strategies applied to all data partitions, which do not account for varying data distributions and attributes, leading to suboptimal performance and resource utilization.

Innovation Solution

Implementing non-homogenous synopsis information and strategies, where the query optimizer dynamically selects the best synopsis strategy based on data partition metadata, such as skewed or uniformly distributed data, to determine which partitions to load into memory, thereby optimizing query execution and resource utilization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a uniform synopsis strategy is applied to all data partitions, then the system structure is simple and easy to implement, but query performance is suboptimal due to inability to adapt to varying data distributions

Engineering Contradiction:
Improveadaptability to varying data distributionsVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies different synopsis strategies to different data partitions based on their specific characteristics. Each partition is evaluated for data distribution properties (skewed vs. uniform) and assigned the most appropriate synopsis strategy accordingly. This local differentiation enables optimal query performance for each partition while maintaining overall system manageability through automated selection mechanisms.

Inventive Principle:
Principle #3Local quality

2Productivity

If more synopsis strategies are implemented to handle different data distributions, then query performance improves, but the complexity of the system increases

Engineering Contradiction:
Improvequery processing efficiencyVSAvoidsynopsis strategy complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system dynamically selects the appropriate synopsis strategy for each data partition based on its data distribution characteristics. Rather than using a fixed uniform strategy, the system adapts its behavior at runtime by evaluating partition properties and choosing the optimal strategy (e.g., histogram-based for skewed data, uniform sampling for uniform data). This dynamic adaptation improves query processing efficiency without requiring manual configuration of multiple strategies.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system automatically determines which synopsis strategy to apply to each partition based on its own data distribution properties. The query optimizer evaluates partition characteristics and self-selects the most appropriate strategy without external intervention. This self-service mechanism enables the system to optimize query performance while managing the complexity of multiple strategies through automated decision-making.

Inventive Principle:
Principle #25Self-service

3Reliability

If all data partitions are loaded into memory for processing, then query accuracy is maintained, but resource utilization is inefficient due to unnecessary data loading

Engineering Contradiction:
Improvequery result accuracyVSAvoidmemory resource utilization
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system performs preliminary evaluation of data partitions using synopsis information before actually loading data into memory for query processing. By pre-evaluating partition properties and selecting appropriate synopsis strategies in advance, the system can accurately predict which partitions are necessary to load and which can be safely pruned. This preliminary action maintains query result accuracy while significantly improving memory resource utilization by avoiding unnecessary data loading.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3518118B1Non-homogenous synopsis for efficient partition pruning
Publication Date: 2023.01.04 SAP SE
  • EP3518118B1 patent drawingFigure 1
  • EP3518118B1 patent drawingFigure 2
  • EP3518118B1 patent drawingFigure 3

AI summary

Disclosed herein are system, method, and computer program product embodiments for partition pruning via non-homogenous synopsis information. An embodiment operates by maintaining synopsis information for a data partition in accordance with a first synopsis strategy, monitoring performance of the synopsis information within a partition pruning system, determining that the performance of the synopsis information is insufficient, and updating the synopsis information in accordance with a second synopsis strategy better suited for the attributes of the data partition. In some embodiments, a first data partition of a partitioned data table may employ a first synopsis strategy and a second data partition of the partitioned data table may employ a second synopsis strategy.