Skip to content

Metric Graphs

LLview's detailed reports visualise each metric using interactive heatmaps (colour maps) that highlight resource usage patterns. See here the full list of metrics.

Info

• X‑axis: either job timestamp or core ID
• Y‑axis: resource dimension (node, GPU, superchip)
• Colour: metric value at each (X,Y) point

Per‑Core Metric Graph

Heatmap of a metric across core IDs (X) and node names (Y)

Time‑Series Metric Graph

Heatmap of a metric over time (X) and node names (Y)

Other examples can be seen here.

Above and beside each heatmap you'll also find three summary curves:

  • Per‑timestamp (top): min, average and max values across all resources at that time.
  • Per‑resource (right): time‑averaged value per node/GPU/superchip, including its min–max range.
  • Global summary (top‑right): overall min, average and max for the entire job.

Info

  • Interactive features:
    • Hover: display exact values and metadata at the pointer.
    • Zoom & Pan: click-and-drag or axis controls to focus on regions of interest.
    • Color-scale dropdowns: choose or reverse the colour map; switch between job-level or system-level bounds.
    • Menubar (hover): copy a direct link to the graph, view its title/description, use zoom/pan buttons, or download the data as JSON.

Use these heatmaps and interactive tools to pinpoint hotspots, compare resources over time, and deep-dive into performance trends.