Interactive QA Dashboard for

LSST Pipelines

Demo #1

• Panel/Holoviews/Bokeh
  • Dashboard framework and interactions
• Datashader
  • Renders aggregated scatter-plot data as image
• Parquet
  • Allows loading only desired columns
• Dask
  • Parallel reading from disk
• Kartothek
  • Smart on-disk partitioning for targeted reads

The Tools

Demo #2

• prepareQADashboard.py

• clone and install dashboard repo & deps

• lsst_data_repartition

• lsst_data_explorer

How to DIY

• Custom/on-demand columns
• Live metric calculations
• Extract selected source IDs
• Drill down to image cutouts
• Save/share state
• Requests?

Future?

(related: any other stack uses of parquet/kartothek partitioning?)