Chart · scatter

Response time against request size

Response time rises with request size in both groups, but the uncached path costs roughly three times as much at any size — and one batch import sits well above its own group.

Response time against request size Illustrative dataset, not a real production trace. Twenty API requests, each measured on two variables: request size in kilobytes on the horizontal axis, zero to 320, and response time in milliseconds on the vertical axis, zero to 600. Both axes include zero, so no correlation is exaggerated by a truncated frame. Ten cached responses, drawn as circles, run from 60 to 185 milliseconds; ten uncached responses, drawn as squares, run from 180 to 545 milliseconds over the same size range. The focal observation is a 272 kilobyte batch import answering in 545 milliseconds, well above the uncached trend. No trend line is drawn, no point is jittered, and no observation was dropped. RESPONSE TIME (MS) · 0–600 600 450 300 150 0 0 80 160 240 320 REQUEST SIZE (KB) · 0–320 Uncached responses Cached responses Batch import — 272 KB, 545 ms

Illustrative dataset — the twenty observations are invented to demonstrate the scatter grammar and describe no real service. Units are stated once per axis, and each axis title prints its full range so the frame is visible. Zero is included on both axes. Groups are told apart by marker shape — circle and square — as well as by token colour, and are labelled directly on the plot rather than through a legend. Marker area encodes nothing; no trend line is fitted. Colour resolves through design-system tokens; where no project tokens are compiled the page falls back to the documented neutral palette declared in :root. Static, script-free, and offline.