Latency
100%

Service Monitoring · Lesson 4

Latency

Learn to compare typical request times with the slowest observations.

Latency is how long a request takes. A single number can hide different experiences. Compare the typical request with the slower ones, using the same route and time window.

Reading a Small Sample

The following five times are in milliseconds. Four requests finished in under 100 milliseconds and one took two seconds.

40  50  60  70  2000

The median, also called the 50th percentile, is the middle value when the sample is sorted. It is 60 milliseconds here. The slowest observation is 2,000 milliseconds. The mean, found by dividing the sum by five, is 444 milliseconds.

Which value describes the middle request in this sorted sample?

Examining the Slow End

A percentile summarizes where a share of observations falls. A 95th-percentile latency, often shown as p95, describes a threshold near the slow end of a larger sample. Use a dashboard's definition consistently. Tiny samples make estimates of the slow end unstable.

The median barely changes, but visitors still report intermittent delays. Which view would help?

Comparing Similar Requests

Compare the same route under similar traffic. A download can reasonably take longer than a small status check. If a slow route stands out, inspect its request logs and dependencies during the same time window. Use the metric to locate the problem and the other evidence to investigate it.

Which comparison best tests whether the same homepage got slower after a release?

Lesson complete

You finished Latency

You can now read typical and slow request times together.

  • Find the median in a small sorted sample.

  • Look at the slow end alongside counts.

  • Compare similar requests before investigating differences.

Keep your learning progress

Create a free account to save this lesson and continue learning on any device.

Create a free account
Next Lesson
Back to Service Monitoring