An observability bill
you can predict.
One price per node, billed on the minutes it ran. Spans, logs, metrics, retention and users are never metered — so nobody has to ration what they collect.
The problem
Most observability bills grow with your data. A price per host per product, then usage on top: gigabytes of logs ingested, millions of lines indexed, spans, custom metrics, longer retention. Each line is reasonable. Together they make next month's bill a guess — and the worst surprises come after an incident, when someone turned on debug logging to find the problem.
So teams ration. They drop logs, shorten retention, sample traces hard, leave APM off the services that seem less important. The savings are real, and so are the blind spots — usually in the place the next incident starts.
One price per node. Nothing else is metered.
$25 per node per month, billed on node-minutes: a node counts while its collector runs and stops the minute it's gone. Everything else is included.
- Spans and traces
- Every request, every service
- Logs
- However many your applications write
- Metrics
- Including custom ones
- Retention
- Keep data as long as your disks allow
- Users
- Everyone who needs to look
- Features
- Profiling, real user monitoring, synthetics, AI investigation
What you pay to run it
Ultraviolet runs in your environment, so you also pay for the infrastructure it uses — to your cloud or your data center, not to us. Mostly that's disk, and it's cheap because telemetry compresses well: a span is stored in about 85 bytes, a typical log line in about 45.
For example, 1,000 spans, 1,000 log lines and 1,000 metric points every second is about 12 GB a day. Kept for 30 days, with room to spare, that's about 520 GB of disk — around $50 a month at typical cloud block-storage prices. Keeping it for 90 days instead costs more disk, and nothing else. The core's CPU and memory and a small collector on each node make up the rest; the sizing guide has the measured figures.
The pricing calculator puts it together for your own numbers — license, disk and compute — next to what the same coverage costs on Datadog at list prices. For a 50-node cluster, Datadog's per-host charges for infrastructure, APM and network monitoring come to $2,550 a month before any log or span charges; Ultraviolet's license is $1,250 (the full comparison).
Levers you control
Retention per type
Keep traces for a week and metrics for a year — you set how long each kind of telemetry is kept.
Log pipeline
Drop health-check spam and debug chatter, or sample noisy sources, before they reach disk — with a dry run that counts what a rule would affect first.
Span sampling
Off by default. If traces outgrow your disk, keep one in N ordinary requests while every error and slow request is kept — request rates, latency, alerts and SLOs stay exact.
It shows your cluster's cost too
The same product shows what your Kubernetes clusters cost and where the money goes — to which namespaces, workloads, teams and node pools — worked out from what it already observes, with no billing access or cloud credentials. It also shows how much you pay for capacity nothing is using, which is usually the largest single saving. Bring in your cloud bill if you like, and it prices nodes at what you were actually charged.
Questions teams ask
- What counts as a node?
- Any Linux machine running an Ultraviolet collector — a Kubernetes node or a standalone host — while its collector is running. Autoscaled, spot and batch nodes cost only the minutes they ran.
- What happens if we go over our committed node count?
- Nothing stops. The product shows the overage and the extra node-minutes are billed at the same per-node rate. Ingest is never blocked.
- A small node costs the same as a large one?
- Yes — one price per node, whatever its size, so the bill follows your fleet rather than your traffic. If your fleet is mostly small nodes, talk to us.
- Is there a minimum?
- Licenses start at 10 nodes ($250 a month). Annual and volume pricing are available.
Try the numbers
Put your node count and traffic into the pricing calculator, or explore the live demo — including its Cost page, on a real cluster. Weighing whether to run an open-source stack yourself instead? The build vs. buy calculator prices the engineers it takes, not just the servers.
Run it in your own cluster this week.
A 30-minute call, then an evaluation core on your infrastructure. You will see your own services, traces and logs — from your own traffic — before you decide anything.
- No data leaves your environment, including during the trial
- Helm install, works fully air-gapped
- You talk to the engineers who built it