Scale Your Infrastructure Without Scaling Your Observability Bill

A Netdata Parent is a centralization point that aggregates observability data from multiple Netdata Agents through real-time streaming. Parents provide unified infrastructure visibility, intelligent workload distribution, and high availability while keeping data processing distributed at the edge. They use the same software as Agents but are configured to accept and maintain data on behalf of other nodes.

A Netdata Parent can handle approximately 200-500 nodes (roughly 1 million metrics per second) with 10 cores and 40GB RAM. For larger deployments, we recommend deploying multiple Parents rather than scaling vertically. This horizontal scaling approach maintains performance and provides natural high availability through active-active clustering.

Agents automatically buffer metrics locally during Parent downtime and replicate historical data when connectivity restores. If you have multiple Parents in an active-active cluster, agents reconnect to available Parents automatically. Netdata Cloud routes queries to available Parents, ensuring continuous visibility. Local agent dashboards remain accessible even during Parent failures.

Agents can offload ML training, health checks, persistent storage, and dashboard serving to Parents. This reduces production system overhead by approximately 50% while maintaining complete visibility. Offloaded agents run in child mode with minimal resource consumption (<2% CPU, <150MB RAM, zero disk I/O). Configuration is simple mode selection - no complex tuning required.

Active-active clustering allows multiple Parents to receive streams from the same agents simultaneously. The first Parent trains ML models; others receive results - providing high availability without duplicating expensive work. Streaming replication between Parents ensures complete data redundancy. Netdata Cloud automatically routes queries to available Parents during failures.

ML models can train at the edge (on agents) or centrally (on parents). When training on parents, the first parent trains 18 unsupervised models per metric; other parents in the cluster receive results. This provides centralized ML management while avoiding duplicate training work. Models flow with metrics stream - train where you have resources, use everywhere.

Storage requirements depend on retention configuration and metrics volume. For 1 million metrics per second with default retention (14 days per-second, 3 months per-minute, 2 years per-hour), expect approximately 2TB storage. Netdata’s industry-leading compression (0.6 bytes per sample for per-second data) enables years of history in gigabytes rather than terabytes.

Yes. Netdata Parents can be deployed entirely on-premises, in your cloud infrastructure, or in hybrid configurations. All observability data stays within your infrastructure - only metadata travels to Netdata Cloud for unified dashboards. This architecture satisfies data sovereignty requirements while providing centralized visibility and team collaboration.

Netdata Parents use push-based streaming with automatic replication, while Prometheus federation uses pull-based scraping. Netdata benchmark testing at 4.6 million metrics/second shows 36% less CPU usage, 88% less RAM consumption, and 97% less disk I/O compared to Prometheus at equivalent scale. Parents include built-in ML anomaly detection, active-active clustering, and zero-configuration agent discovery - capabilities that require external solutions in Prometheus deployments.

Netdata uses predictable per-node pricing with volume discounts that covers agents, parents, and all features. There are no separate charges for parents, no per-metric fees, no ingestion costs, and no overage charges. This pricing model scales linearly with infrastructure growth rather than exponentially with data volume. See our pricing page for current details.

Install Netdata Agents on your infrastructure using our one-line installer. Deploy Parents on dedicated servers or VMs. Configure agents to stream to parents by setting parent endpoints in stream.conf. Parents auto-discover streaming agents - no manual registration required. For detailed guidance, see our Parents documentation or contact our team for architecture consultation.