Netdata

What it does

Netdata collects infrastructure and application metrics at one-second resolution and builds dashboards for them automatically. Install the agent and charts appear — no scrape configs, no dashboard JSON, no deciding in advance which metrics matter.

That resolution is the point. Most monitoring averages to 10, 30 or 60 second intervals, and a CPU spike or a lock contention event lasting four seconds disappears into the mean. Netdata keeps it visible.

Collection model

The agent runs on each node, auto-discovers what is running, and starts collecting. Metrics stay local by default, which keeps egress down and means the dashboard works even when the cloud layer is unreachable. Netdata Cloud sits above the agents to give one view across a fleet.

Licensing and cost

The agent is GPLv3+ and free. Read the tier conditions carefully, though: the free Community tier is limited to personal, non-commercial use and caps you at 5 connected nodes and 1 custom dashboard per Room. Commercial use needs a paid plan.

Homelab is $90/year for unlimited nodes under a fair-use policy. Business is $4.50 per node per month billed annually, adding unlimited dashboards, RBAC and SSO, with a 14-day trial. Enterprise On-Premise is quoted by sales and starts at 200 node licences.

Where the model strains

Node-based pricing works out well for a few dozen machines and badly for a few thousand — at $4.50 a node, a 1,000-node fleet is $54,000 a year, which is where per-GB competitors start looking cheaper.

Per-second data is also expensive to keep. Netdata's default is generous local retention with downsampling over time, so historical analysis loses the resolution that made the live view useful.

Netdata is strongest as infrastructure and system monitoring. For request-level distributed tracing across services, an OpenTelemetry backend fits the job better.

Deployment shape

Netdata is agent-plus-parent rather than agent-plus-SaaS. Agents can stream to a parent node that aggregates and retains data for a group of machines, which keeps metrics inside your network and reduces the number of machines holding long retention. Netdata Cloud then reads from the parents.

This matters for two reasons. Air-gapped or regulated environments can run the whole thing locally, with Cloud as an optional view rather than a dependency. And because each agent serves its own dashboard, a node under investigation can be inspected directly even when the aggregation layer is down — the opposite of a SaaS-only tool, which goes blind exactly when the network is in trouble.

Reading the alerts

Netdata ships a large set of default alarms, and they fire more readily than most teams expect given per-second data. Tuning the alarm set is part of setup rather than an optimisation to do later; left at defaults on a busy host, the notification volume trains people to ignore it.