Box Plot for Server Latency Percentiles — Microservice P25, P50 & P75
A server latency box plot compares response-time quartiles ($P_{25}, P_{50}, P_{75}$) and upper tail whiskers across distributed endpoints to identify high-jitter services.
Tip: Paste Excel or Google Sheets rows (Ctrl+V / ⌘V) directly into any cell.
| # | Label | ||||
|---|---|---|---|---|---|
| 1 | |||||
| 2 | |||||
| 3 | |||||
| 4 | |||||
| 5 | |||||
| 6 |
Latency Jitter (IQR) = P75 - P25. Relative dispersion = (P75 - P25) / P50.
Chart Design & Use-Case Tips
- 01Latency Jitter (IQR) = P75 - P25. Relative dispersion = (P75 - P25) / P50.
- 02A tall box (large gap between P25 and P75) indicates high latency jitter caused by lock contention, GC pauses, or noisy-neighbor I/O.
- 03Compare regional deployments (us-east-1, eu-central-1, ap-northeast-1) side-by-side before and after database read-replica migrations.
- 04Use in SRE architecture reviews and quarterly reliability reports.
Frequently Asked Questions
What does a wide IQR box on a microservice indicate?+
A wide interquartile box means latency is unpredictable—even if the median (P50) looks acceptable, a large fraction of requests experience 2x–3x slower execution.
Can I enter P90 or P95 in the third column instead of P75?+
Yes—many SRE teams enter P50 in Column 1, P25 in Column 2, and P90 or P95 in Column 3 to visualize the 25th-to-95th percentile envelope.
How does this complement the API Latency Histogram?+
Use the API Latency Histogram to inspect the full bin-by-bin curve of a single endpoint, and use this Box Plot to compare 5 to 10 endpoints side-by-side.