Your own Grafana stack. One flat bill.
Deploy the complete Grafana observability stack in your own cloud. Metrics, logs, traces, and profiles, correlated and governed by you. Flat pricing. No surprises.
One click, five stops, none of them ours.
No jargon required. Scroll through what actually happens between someone clicking a button in your product and that click showing up on the dashboard above.
Something happens, and your app writes it down
A customer clicks Buy. As your app handles that click, it quietly jots down a note: what was asked for, how long it took, and whether it worked. You don't write this code. It's added for you.
OpenTelemetry SDK auto-instrumentation. Each note is a span on a trace.
The notes are bundled up and mailed
Sending each note the instant it's written would be like driving to the post office for every single letter. Instead your app fills a bag, and ships the whole bag at once. Cheaper, faster, less noise.
Spans batch in-process (512 or 5s, whichever first), then leave over OTLP to a collector in your VPC.
Everything gets counted. Only the interesting parts get kept
This is the step that keeps the bill flat. Every single note is counted first, so your totals are never guesses. Then the boring ones get thrown away and only the useful ones are stored: anything that broke, anything that was slow, plus a small random sample of the rest.
Tempo's metrics-generator computes RED metrics from 100% of spans, then tail-sampling keeps errors and slow requests plus a small baseline of the rest.
It lands in your filing cabinets, in your building
Four kinds of data, four engines built to hold them, each with its own retention. Metrics keep 427 days, logs 97, profiles 30, traces 14. All four write to buckets inside the cloud account you already own. We never hold a copy.
Mimir (metrics, Kafka-backed ingest), Loki (logs, zone-aware ingesters + bloom filters), Tempo (traces), Pyroscope (profiles), all writing to your own S3/GCS.
You ask questions and get answers instantly
Your dashboard asks those cabinets questions, constantly: how fast are we, what's broken, is this worse than an hour ago. The answers come back in milliseconds, and the question language is an open standard, so it still works anywhere else.
Grafana queries storage directly via PromQL, LogQL, and TraceQL. No proprietary layer between.