Parallel Computing Visualization
summary
A live visualization for multi-process builds shows real data: the free execution visualization tool (incredible) maps tasks to cores, colors distributable tasks green and local-only blue, and marks warnings and errors yellow and red, with double-click jumps to textual output and replayable traces. It compresses and shareably exports logs, saves to a database, and claims dramatic wins — for one example reduced a 15 minute build to 1.4 minutes; a freemium Visual Studio 2017 edition runs locally up to 16 cores.
AI-written summary · about
chapters (21)
- 0:00Introduction and purpose
- 1:46Parallel computing challenges
- 4:53Incredibuild solution overview
- 8:26Visualization monitor features
- 13:27Zooming and data scaling
- 15:47Contextual data in visualization
- 17:24Analyzing build bottlenecks
- 19:45Predictive features and scaling
- 22:30Resource utilization analysis
- 24:49Tool hints and searching
- 28:20Graph correlation techniques
- 31:10Extended logging and R&D
- 34:23Task dependency analysis
- 37:21Low-level task diagnostics
- 38:45Agent-level visualization
- 40:42Database archiving and replay
- 42:51Attachments and tool availability
- 45:11Summary and call to action
- 49:33Handling large data volumes
- 52:01Q&A and Linux support
- 56:25Closing remarks
related talks
21mGraphing tools for scheduler tracing
44mNow You See Me Too: Visual Tooling for Advanced System Analysis
46mExplorations in Interactive Visual Analytics
08m#BuildPropulsionLab at #DevoxxUK: @jbaruch with @advarave
34mAdvanced visualization of Flink and Spark jobs
01h 07mVisualizing Connections
37mVisualizing Persistent Data Structures
39mSparklens: understanding the scalability limits of Spark applications