Realtime metrics
Counters, histograms and gauges arrive over a single agent and render while they land. Pin the graphs a team argues about to a board and everyone reads the same numbers.
ingest lag 1.4 sInfrastructure visibility without the noise.
Monitor services, latency and deployments from one clean workspace.
Edge latency, last 6 hours
p99 across 22 probes
p99
p50
Bucket 30 s
checkout-api38 ms
ledger-worker64 ms
media-transcode417 ms
search-index91 ms
Start with one service
Bring one service across and see whether the timelines hold up under a real incident.
Where the probes run
A region leaves this table after two consecutive failed sweeps. Longer histories stay available in the workspace.
| Region | Checks/min | p50 | p99 | Status |
|---|---|---|---|---|
| us-east-1 | 1 840 | 28 ms | 41 ms | Normal |
| us-west-2 | 1 220 | 33 ms | 57 ms | Normal |
| eu-west-2 | 1 465 | 36 ms | 58 ms | Normal |
| eu-central-1 | 980 | 39 ms | 74 ms | Elevated |
| ap-south-1 | 610 | 71 ms | 112 ms | Normal |
Incident timelines
Every alert opens a timeline that collects deploys, config changes and acknowledgements in order. Nothing has to be reconstructed from chat afterwards.
median ack 4 m 12 sGlobal checks
HTTP, TCP and DNS probes run from twenty-two cities on a fixed interval. A check has to fail in two locations before it pages anyone.
22 cities, 60 s interval