Cut your observability bill by 90%

Data size is simple and predictable. An average span is about 500 bytes plus custom attributes, so you can estimate costs easily.

This model is fairer for microservices that generate many small spans and RPC calls. You pay for what you actually store, not arbitrary span counts.

Metrics are billed per million ingested datapoints. Typically, each timeseries produces one datapoint per minute, so 1,000 timeseries over 28 days is about 40 million datapoints (1,000 * 28 * 24 * 60).

The exact number of datapoints depends on collection_interval in OpenTelemetry Collector or scrape_interval in Prometheus (both default to 1 minute). For example, with collection_interval=30s the number of datapoints per minute will double.

Datapoint-based billing lets you use different collection intervals for different metrics — collecting critical metrics more frequently while keeping others at the default interval. The minimum supported interval in Uptrace is 10 seconds.

A timeseries is a metric with a unique set of labels. For example, these are 3 separate timeseries:

• cpu_usage{host="host1"}
• cpu_usage{host="host2"}
• cpu_usage{host="host3"}

The rate starts at $0.025 per 1M datapoints and decreases with volume. See the Metrics tab above for the full tier table.

Uptrace Cloud keeps all data for 4 weeks (28 days) by default.

Need longer retention? Store data on cold storage (S3 or HDD) for up to 100 weeks:

Example: retaining 1,000 GB of spans beyond 4 weeks costs $10/month.

We use Paddle for secure billing and global tax compliance. Your payment details stay with Paddle, so we never see them. Receipts are handled by Paddle.

Pay by credit card (automatic) or bank transfer (manual).

Yes. Configure Uptrace to sample a percentage of data and drop the rest to lower your bill.

Dropped data costs just $0.01 per GB. For example, dropping 1,000 GB costs $10. No sampling? No charge.

Uptrace Cloud runs in Hetzner's Germany datacenters with redundant backups in Finland. See our privacy policy for details.

Sending from AWS or GCP? Outbound fees apply, but observability data compresses well. OTel Arrow cuts bandwidth 30–70% versus standard OTLP.

Yes. Install Uptrace on-premises with a license. Your telemetry never leaves your infrastructure, making it ideal for compliance and security audits.

On-premises includes:

• Free evaluation period
• Terraform and Ansible automation
• Custom retention policies
• Dedicated support engineer
• SOC 2 certification on request

Yes. Uptrace is built on ClickHouse, a columnar database designed for petabyte-scale analytics. Customers run Uptrace in production with billions of spans per day.

Both Uptrace Cloud and on-premises deployments support horizontal scaling, so you can add capacity as your traffic grows without re-architecting.

ClickHouse handles high-cardinality attributes (user IDs, request IDs, container names) efficiently thanks to columnar compression and sparse indexing.

Unlike time-series databases that choke on high-cardinality labels, Uptrace stores spans and logs as structured events, so adding more unique attribute values does not degrade query performance.

Nothing breaks. You keep sending data and Uptrace keeps ingesting it. At the end of the billing cycle, any usage above your plan is charged at the standard per-GB rate.

You can set budget alerts to get notified before you reach your cap, and you can always adjust your plan mid-cycle.

Yes. Explore our public live demo. No account needed.

Yes, two options. Self-host Uptrace Community Edition for free forever with no limits.

Or use Uptrace Cloud free up to 50 GB and 5,000 timeseries per month.

Email us or schedule a call. We typically respond within a few hours.