Data Wrangling Execution Engine Consistency via Local Sampling

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Data wrangling operations on small datasets are slower and inconsistent across different execution engines, leading to varying outputs even when the same operations are performed, due to the lack of guarantee in behavior between engines suitable for large and small datasets.

Innovation Solution

Implementing a system where a local client device plans and previews data wrangling operations, transferring a selected portion of the dataset for local processing using the same execution engine as the remote device, ensuring consistent output and reducing the need for multiple execution engine implementations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If different execution engines are used for large and small datasets, then processing efficiency is improved, but behavior consistency deteriorates

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidbehavior consistency
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system segments the execution engine functionality into a standardized interface layer and implementation layer. The standardized interface ensures consistent behavior across different data sizes, while allowing optimized implementations for specific scenarios. This segmentation resolves the contradiction by maintaining uniform behavior contracts while enabling efficiency optimizations.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a universal execution engine design that can handle both large and small datasets through a common interface. This multi-functional approach allows the same execution engine to adapt to different data scales while maintaining consistent behavior, eliminating the need for separate engines and their associated inconsistency problems.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If multiple execution engine implementations are maintained, then adaptability to different data sizes is improved, but device complexity increases

Engineering Contradiction:
Improveadaptability to data sizeVSAvoidexecution engine management complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system employs a universal execution engine interface that can adapt to different data sizes without requiring multiple separate implementations. This single interface handles both small and large datasets, significantly reducing the complexity of managing multiple engines while maintaining full adaptability.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent introduces an intermediary layer (the standardized execution engine interface) that mediates between the user code and the actual data processing operations. This intermediary abstracts away the complexity of different engine implementations, allowing users to write code once while the system adapts to different data scales behind the scenes.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Loss of time

If data wrangling operations are performed locally on small samples, then preview and planning efficiency is improved, but data representation accuracy may deteriorate

Engineering Contradiction:
Improvepreview planning timeVSAvoiddata representation accuracy
Core Design Contradiction:
Loss of timeVSMeasurement precision

Solution Approach 1:

The system performs preliminary data wrangling operations on local samples to preview and plan the full data processing pipeline. This preliminary action allows users to validate their wrangling logic on a smaller scale before committing to full dataset processing, saving time while maintaining accuracy through the standardized execution engine interface that ensures consistent behavior across different data sizes.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3040879B1Behaviorally consistent cluster-wide data wrangling based on locally processed sampled data
Publication Date: 2021.11.03 BUSINESS OBJECTS SOFTWARE
  • EP3040879B1 patent drawingFigure 1
  • EP3040879B1 patent drawingFigure 2
  • EP3040879B1 patent drawingFigure 3

AI summary

Example embodiments involve a system, computer-readable storage medium storing at least one program, and computer-implemented method for behaviorally consistent data wrangling. A local client device selects a set of raw sample data from a remote datastore. A local execution engine then applies one or more local data wrangling operations to the raw sample data. If the results of the local data wrangling operations are satisfactory, the local data wrangling operations may then be transferred to a remote data wrangling cluster. A remote execution engine being executed by the remote data wrangling cluster then applies the data wrangling operations to the larger set of raw data from which the sample raw data was obtained. As the remote execution engine and the local execution engine are of the same type, the data wrangling behavior exhibited by the local execution engine is reflected in the data wrangling behavior of the remote execution engine.