Compute Workflow Visualization with Multi-Level Abstraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing compute workflow visualization tools fail to display multiple levels of abstraction, making it difficult for operators to understand performance and compare workflows effectively, as they typically present only one level of abstraction at a time and require a click-through paradigm.
Innovation Solution
A method and system for visualizing compute workflows by obtaining workflow and job data, generating a visual representation of resource utilization over time for each computation stage, and rendering multiple instances on a common timeline, using a two-dimensional coordinate system with tasks represented as shapes corresponding to duration and resource usage, and transforming these shapes into polygons for integrated viewing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If multiple levels of abstraction are displayed simultaneously, then workflow performance understanding is improved, but device complexity increases
Solution Approach 1:
The visualization system segments workflow information into distinct hierarchical levels (workflow level, computation stage level, task level), each represented by separate graphical elements. This allows operators to view multiple levels simultaneously without overwhelming complexity, as each level can be independently analyzed while maintaining contextual relationships through the hierarchical structure.
Solution Approach 2:
The patent introduces a hierarchical dimension to the visualization, organizing graphical elements across multiple levels of abstraction. By stacking elements vertically to represent different hierarchical levels while maintaining temporal progression horizontally, the system displays comprehensive workflow information without increasing horizontal complexity, effectively using vertical space to manage information density.
2Device complexity
If a click-through paradigm is used to navigate different abstraction levels, then device complexity is reduced, but loss of information increases
Solution Approach 1:
The visualization system merges multiple levels of abstraction into a single integrated display rather than requiring separate views. All hierarchical levels (workflow, stage, task) are presented simultaneously in one visualization, eliminating the need for click-through navigation and preserving complete workflow context in a unified view.
Solution Approach 2:
By adding a vertical hierarchical dimension to the visualization, the system accommodates multiple levels of abstraction within a single view. This dimensional approach allows operators to see all relevant information simultaneously without requiring multiple screens or click-through interactions, maintaining context while managing visual complexity.
3Measurement precision
If detailed task-level information is displayed for all workflow instances, then measurement precision is improved, but data processing time increases
Solution Approach 1:
The system segments the display of task-level information by hierarchical level, showing aggregated metrics at higher levels and detailed task information at lower levels. This selective segmentation allows precise task-level measurement where needed while avoiding the processing overhead of displaying all detailed information uniformly across all workflow instances.
Solution Approach 2:
The visualization system applies partial detail display by showing comprehensive task-level information only for selected or anomalous workflow instances rather than all instances. This partial action approach maintains measurement precision for analysis purposes while reducing overall data processing and rendering time by selectively applying detailed visualization.
Data Source
AI summary
A system and method for visualization of a compute workflow are provided. Workflow data indicative of one or more jobs executed in each computation stage of each instance of the compute workflow and job data indicative of one or more tasks executed by each of the one or more jobs are obtained. Each of the one or more tasks is associated with a corresponding computation stage. For each instance of the compute workflow, a visual representation of a resource utilization over time of each of the one or more tasks executed in each computation stage is generated. The visual representation is rendered for multiple instances of the compute workflow on a common timeline.