Datadog infrastructure billing covers how Datadog meters Infrastructure Monitoring and the products tied to it: containers, custom metrics, custom events, AWS Fargate tasks and serverless functions. The base unit is the host, which Datadog defines as any physical or virtual operating system instance that it monitors. A host can be a server, a VM, a Kubernetes node, an Azure App Service Plan instance or a Heroku dyno.[1] Each paid host licence comes with allotments of containers, custom metrics and custom events. Usage beyond those allotments is billed through separate SKUs.[3] The overview is in Datadog licensing.
Editions
Infrastructure Monitoring has a Free tier and four paid editions:[3][4]
| Edition | Annual / month-to-month / on-demand per host (USD) | Containers per host | Custom metrics per host | Custom events per host |
|---|---|---|---|---|
| Free | 0 (up to 5 hosts, 1-day retention) | |||
| Infrastructure Pro | 15 / 18 / 18 | 5 | 100 | 500 |
| Infrastructure Enterprise | 23 / 27 / 27 | 10 | 200 | 1,000 |
| DevSecOps Pro | 22 / 27 / 27 | 5 | 100 | 500 |
| DevSecOps Enterprise | 34 / 41 / 41 | 10 | 200 | 1,000 |
The allotment figures come from the pricing page’s feature table and Datadog’s allotments table.[3][12] Enterprise adds features such as machine-learning alerts, Live Processes and the Governance Console.[3] The DevSecOps editions add Cloud Security Management. A DevSecOps SKU cannot be bought for only part of the infrastructure: it covers all existing hosts and containers. It also cannot be combined with standalone Infrastructure or Cloud Security SKUs.[3] Catalog: DevSecOps SKUs must cover all hosts and cannot be combined with standalone SKUs.
Several items are priced separately on the price list. Container Monitoring costs USD 1 per prepaid container per month, or USD 0.002 per container hour on demand. Custom Metrics cost USD 5 per 100 per month, Ingested Custom Metrics USD 0.10 per 100, and Custom Events USD 2 per 100,000 when billed annually.[4]
Metrics
| Unit | Catalog row | Definition |
|---|---|---|
| Host | Host | Physical or virtual OS instance; Datadog records unique hosts each hour.[2] |
| Container | Container | Self-contained operating environment with application software and limited OS libraries and settings; recorded every five minutes.[2] |
| Custom metric | Custom metric | Identified by a metric name plus tag values, including the host tag.[6] |
| Metric name | Metric name | Under Metric Name pricing, each unique metric name with more than 100 indexed datapoints in a month.[7] |
| Fargate task | Fargate task | An ECS task; ECS Fargate counts concurrent tasks and EKS Fargate counts concurrent pods.[2][5] |
| Active function | Active function | A unique function ARN executed at least once in the hour.[11] |
Counting / floors
Hosts
Datadog meters hosts every hour. At the end of the month, the billable count is the highest hourly count in the lower 99 percent of hours. The top 1 percent of hours is excluded to reduce the effect of spikes.[1] In a 30-day month (720 hours), this means the seven busiest hours are ignored. The documentation calls this the high watermark plan. Under the alternative hybrid monthly/hourly plan, the customer pays a monthly minimum commitment, and every host hour above it is charged at an hourly rate.[2] Catalog: Billable host count is the high-water mark of the lower 99 percent of hours; Hybrid monthly/hourly plan bills host hours above the monthly commitment hourly.
Hosts that stop reporting (status INACTIVE) do not count towards billing. They can take up to two hours to drop out of the Infrastructure List.[1] A licence manager comparing an inventory with the Datadog bill should therefore compare peak concurrent hosts, not the number of hosts that existed at some point in the month. Catalog: Non-reporting hosts do not count towards billing.
Containers
Datadog recommends one containerized Agent per host, which collects both container and host metrics. If the Agent is installed inside each container, each container is billed as a host.[1] Containers come with Pro and Enterprise plans: 5 or 10 per host licence, averaged across the whole infrastructure.[5] Because the allotment is pooled, unevenly spread containers cost nothing extra as long as the total stays within the account’s allotment.[5]
Containers above the allotment are metered in five-minute increments. Each hour’s samples are averaged by dividing by twelve, the allotment is deducted hour by hour, and the hourly on-demand amounts are added up for the month. A burst of 1,200 extra containers in a single five-minute interval therefore counts as 100 container hours.[5] The account allotment is the sum of the included containers and any contracted container commitment. Prepaid containers can also be bought.[5] Kubernetes pause containers are excluded, as are the Datadog Agent’s own containers.[5] A container in a pod stuck in CrashLoopBackoff counts if it runs for more than ten seconds in a metering interval.[5] Catalog: Container allotment is 5 or 10 per host licence, averaged across the infrastructure; On-demand containers are metered in five-minute increments and averaged per hour; Datadog Agent containers and Kubernetes pause containers are not counted.
Custom metrics
Any metric not submitted by one of Datadog’s integrations is a custom metric, and some standard integrations can also emit them. A custom metric is uniquely identified by its metric name and tag values, including the host tag.[6] Billable usage is the total of distinct custom metrics (timeseries) in each hour of the month, divided by the number of hours in the month. How often data points are submitted, and how often the metrics are queried, makes no difference.[6] The allotment is 100 custom metrics per Infrastructure Pro host or 200 per Enterprise host. It is counted hourly and averaged monthly across the account, not per host.[3] Each tag value multiplies the number of timeseries, so tag cardinality is the main cost driver. Catalog: Custom metrics are billed on the monthly average of hourly distinct timeseries; Custom metric allotment is 100 per Pro host and 200 per Enterprise host.
With Metrics without Limits, a customer chooses which tags stay queryable. Configured metrics then report two volumes: ingested custom metrics (all tags sent) and indexed custom metrics (the tags kept). Ingested custom metrics above the allotment cost USD 0.10 per 100, and indexed custom metrics are billed at the contract rate.[6] Metrics not configured this way are billed only on indexed volume.[6]
Some contracts use Metric Name pricing instead. It bills each unique metric name in the month that has more than 100 indexed datapoints. Each billed name includes 10 million indexed datapoints, and ingested datapoints above five times the indexed volume are billed separately.[7] Rates fall in five marginal tiers. Month-to-month rates are 20% above the annual baseline and on-demand rates 40% above it. Metric Name SKUs cannot be combined with the timeseries SKUs.[7] Catalog: Metric Name pricing SKUs are mutually exclusive with timeseries pricing.
Custom events
Pro hosts include 500 custom events and Enterprise hosts 1,000, averaged across the account. Above the allotment, Datadog charges USD 2 per 100,000 custom events billed annually, or USD 3 on demand.[3] Catalog: Custom events are allotted per host and averaged across the account.
Serverless and Fargate
Serverless monitoring is billed on the average number of active functions per hour across the month. Each hour, Datadog records the functions executed at least once. A billable function is a unique function ARN, and Lambda@Edge functions count separately in each region.[11] Pro and Enterprise plans include five custom metrics per billable function.[11] Functions instrumented with Datadog’s Lambda libraries are billable even if the AWS integration is disabled.[11] Fargate is charged on concurrent monitored ECS tasks and EKS pods.[5] GKE Autopilot environments are billed under the Serverless category rather than as hosts.[5] Catalog: Serverless functions are billed on the monthly average of hourly active functions.
Floors. The price list sets no minimum host quantity.[4] Minimums come from committed quantities in an Order. Volume discounts are offered from 500 hosts per month.[3]
Virtualization & partitioning
Datadog has no hypervisor or partitioning rule. Each VM and each Kubernetes node is a host.[1] For vSphere, Datadog bills each Agent installed on a vCenter server and each VM and ESXi host monitored. VMs can be filtered out of monitoring with a regular expression in the integration configuration.[13] For APM, Kubernetes is priced by node, not by pod. That rule is covered in Datadog APM and Log Management billing. See virtualization and partitioning for the general concept.
Cloud / BYOL
The cloud integrations change the host count because they discover instances without an Agent:
- AWS. Datadog bills hosts running the Agent and all EC2 instances picked up by the AWS integration, without billing an instance twice.[8] Only EC2 (hosts), Lambda (active functions), CloudWatch custom metrics and containers are billable. Other services such as ELB, RDS and DynamoDB do not incur Datadog charges.[8] If instances require IMDSv2, Agent 7.64.0 or later, or the
ec2_prefer_imdsv2setting, is needed to avoid double billing.[8] Catalog: Only EC2, Lambda, CloudWatch custom metrics and containers are billable from the AWS integration. - Azure. All monitored Azure VMs are billable whether or not the Agent is installed. Nodes inside App Service Plans also count as hosts, but Shared, Dynamic and Free tier plans do not.[9] Catalog: Azure VMs are billable whether or not the Agent is installed.
- Google Cloud. Datadog bills hosts running the Agent and all GCE instances picked up by the integration. Services outside GCE, such as Dataflow, can create billable GCE hosts.[10]
Each integration has tag-based filters that limit which instances are collected. Filtered EC2 instances can stay in the Infrastructure List for up to two hours, but they are not billed during that period.[8] Turning on collection of custom metrics from Azure App Insights or CloudWatch adds to custom metric usage.[9][8]
Programs
Infrastructure products can be bought at annual, month-to-month or on-demand rates, and commitments can be monthly or hourly.[4][3] How commitments, allotments and on-demand usage combine is described in Datadog commitments, allotments and subscription terms.
Out of scope
- Network Device Monitoring, Cloud Network Monitoring, NetFlow, GPU and Edge Device Monitoring. They have their own units on the price list.
- Cloud Security Management, Workload Protection and other security products sold separately from the DevSecOps bundles.
- Datadog for Government sites, where some products are not available.