Solutions
Follow a playback slowdown across service dependencies.
A streaming service sees viewers waiting longer for video to start. One latency alert gives the on-call engineer a request path to follow instead of a list of suspects.
The situation
It is early evening, the busiest hour for a streaming service. Support has not heard anything yet, but the playback service's latency has started to climb, and an alert has fired for the service that hands video to viewers' devices.
Nothing is down. Requests are still succeeding, just more slowly, which is the kind of problem that is easy to see on a chart and hard to explain from one.
What the on-call engineer sees first
Service health shows the playback endpoint turning amber while its neighbours stay green. The alert carries the time the change began, so the engineer can set the same window on every other view and stop guessing when it started.
The first useful fact is a negative one: the front door is fine. The slowdown is somewhere behind it.
How the investigation goes
The engineer opens a handful of slow requests from that window and reads their traces. In each one the time is not spent in the playback service itself but in a single call it makes to a service that prepares the stream's manifest.
The dependency map confirms it: the manifest service sits behind playback, and its own downstream call to a catalogue lookup is the widest span in every slow trace. Logs from the same requests and the same minutes show that lookup retrying.
That is a path, not a hunch. The engineer can say which hop is slow, since when, and what it is waiting for.
What gets shared
Instead of pasting screenshots into a chat, the engineer puts the time window, two representative traces and the affected services into the incident workspace. The team that owns the catalogue lookup opens the same evidence and starts from the slow call, not from a summary of it.
What this shows
A latency alert is only the start of a question. The value is in moving, in a few clicks, from the symptom viewers feel to the service that causes it, with the same evidence in front of everyone who has to act.
These stories are illustrative examples. They describe invented scenarios in an industry, not a real customer, a real incident or any company's systems.
Start with your own services.
A trial workspace is free for 14 days. Send your telemetry and follow your own requests.