
What it tells you
- Total request volume in the selected window.
- How that compares to the previous period.
- Hourly / daily distribution — peaks, troughs, anomalies.
How to read the chart
| Element | Meaning |
|---|---|
| Main number | Total requests (M / B suffix). |
| Percentage | Delta vs the prior identical window. |
| Sparkline | Time-series — find spikes and dips. |
Pull via API
Requests vs traffic
A high request count with low traffic means tiny payloads (API JSON, manifests). A high traffic with low requests means big objects (video, downloads). Watch both ratios.| Ratio | Workload |
|---|---|
| Many requests, low bytes/req | API-style or HLS manifest. |
| Few requests, high bytes/req | Video / downloads / large images. |
| Both high | Mixed-purpose distribution — scope per-path metrics. |
Spot abuse
Sudden request spikes without a corresponding traffic spike usually mean:- Bot scraping — same small object hit thousands of times.
- Cache busting — random query strings forcing every request to be unique.
- Probing —
404flood from a vulnerability scanner.
Operational tips
- Group by path. The top 10 paths by request count usually drive the bulk of origin traffic.
- Watch off-hours. Sustained traffic at 03:00 local is rarely human.
- Combine with cache-hit ratio. High request count + low HIT% = origin pain. Tune cache rules.
- Alert on rate change, not raw totals — easier to write portable thresholds.