Dynatrace host and Kubernetes monitoring consumption describes how the Dynatrace Platform Subscription turns monitoring into billable units. Monitoring a host or container is billed according to the OneAgent monitoring mode chosen for it. Full-Stack Monitoring is charged on memory, while Infrastructure Monitoring and Foundation & Discovery are charged per host regardless of size.[6] Kubernetes, code monitoring, logs, real user monitoring and application security have their own units. All of these are drawn down from the same annual commitment. The overview is in Dynatrace licensing.
Editions
The three host monitoring modes form a ladder:[12][6]
| Mode (catalog SKU) | Scope | Unit | List price |
|---|---|---|---|
| Foundation & Discovery | Basic host health, host and process topology, disk and OS service status; logs in context | Host-hour, any host size | USD 0.01 per hour (USD 7 per month) |
| Infrastructure Monitoring | Adds process, disk, memory and network analysis and custom metrics | Host-hour, any host size | USD 0.04 per hour (USD 29 per month) |
| Full-Stack Monitoring | Adds APM, code-level profiling, distributed tracing and Kubernetes Platform Monitoring | Memory-GiB-hour | USD 0.01 per GiB-hour (USD 58 per month for an 8 GiB host) |
Dynatrace positions Foundation & Discovery for broad deployment on all hosts, to decide where deeper monitoring is worth it.[3] Full-Stack Monitoring also comes as container-based application-only monitoring, which is charged on container memory. That variant does not include Infrastructure Monitoring or Kubernetes Platform Monitoring.[1]
Metrics
| Unit | Catalog row | Definition |
|---|---|---|
| Memory-GiB-hour | Memory-GiB-hour | GiB of monitored host or container memory per hour; more memory and longer monitoring consume more.[1] |
| Host-hour | Host-hour | One hour of active monitoring of one host, an OS instance with OneAgent installed and running.[2] |
| Pod-hour | Pod-hour | One hour of monitoring of one Kubernetes pod.[4] |
| Container-hour | Container-hour | Code Monitoring unit; each concurrent instrumented process or container counts.[5] |
| GiB; GiB-day; GiB scanned | GiB ingested; GiB-day; GiB scanned | Raw data sent before enrichment; data stored before compression, per day; data read by a DQL query.[7][8][9] |
| Session | RUM session | One user’s visit to an application.[10] |
Counting / floors
Full-Stack Monitoring
Each OneAgent in Full-Stack mode consumes GiB-hours based on the host’s physical or virtual RAM, measured in 15-minute intervals. A host monitored for less than 15 minutes in an interval is counted for the full 15 minutes.[1] RAM is rounded up to the next 0.25 GiB, and every physical or virtual host counts as at least 4 GiB. An 8.3 GiB host therefore counts as 8.5 GiB, and a 2 GiB host as 4 GiB.[1] Technology factors such as the number of JVMs or microservices do not affect consumption.[1] Catalog: Full-Stack hosts are measured by RAM rounded up to 0.25 GiB with a 4 GiB minimum; Technology factors such as JVM count do not affect Full-Stack consumption.
Application-only container monitoring uses the same 0.25 GiB rounding, but the minimum is 256 MiB rather than 4 GiB. A 780 MiB container counts as 1 GiB.[1] Recent OneAgent versions measure the container’s used memory (1.275+ for Kubernetes, 1.297+ for serverless containers). Older versions use the container memory limit, or the underlying VM’s memory if no limit is set.[1] Some scenarios have their own rules. Azure App Service on Windows dedicated plans is counted by plan instances and their memory, with a 256 MiB minimum. Oracle Solaris Zones are counted as hosts, and so are containers that Dynatrace does not detect as containers.[1] Catalog: Application-only container monitoring uses container memory with a 256 MiB minimum; Solaris Zones and undetected containers are counted as hosts.
Included volumes. Full-Stack Monitoring includes all built-in infrastructure and APM metrics at no charge. Each contributing GiB of memory also adds 900 custom metric data points per 15-minute interval.[1] Each contributing GiB adds 200 KiB of trace data per minute, or 3,000 KiB per 15-minute interval. Environments below 320 contributing GiB get a fixed minimum peak volume instead.[1] Unused included volume does not roll over to later intervals. Data points above the included volume are charged as Metrics powered by Grail, or as Custom Metrics Classic if Grail metrics are not on the rate card.[1] The first 10 days of trace retention are included. Longer retention is charged as Traces - Retain.[1] Catalog: Full-Stack includes 900 custom metric data points per GiB per 15 minutes; Full-Stack includes 200 KiB of trace data per GiB per minute.
Infrastructure Monitoring and Foundation & Discovery
Both modes are measured in host-hours. Each monitored host contributes one host-hour per hour, whatever its RAM, with the same 15-minute rounding.[2][3] Infrastructure Monitoring includes 1,500 custom metric data points per host in each 15-minute interval, and unused points do not roll over.[2] Foundation & Discovery cannot ingest custom metrics. Logs it collects consume Log Management and Analytics.[3] Dynatrace’s examples, at list price, for 100 hosts monitored around the clock for 30 days: USD 2,880 per month on Infrastructure Monitoring and USD 720 per month on Foundation & Discovery.[2][3] Because these modes charge per host rather than per GiB, large hosts cost relatively less than on Full-Stack. At list prices, a host of 4 GiB or less costs the same per hour on Full-Stack as on Infrastructure Monitoring (USD 0.04), because of the 4 GiB minimum. Any larger host costs more on Full-Stack. Catalog: Infrastructure Monitoring is one host-hour per host regardless of memory; Foundation & Discovery does not include custom metric ingest.
Kubernetes and containers
Kubernetes Platform Monitoring charges one pod-hour per monitored pod per hour, regardless of the pod’s memory, CPU or number of containers. Billing uses 15-minute intervals.[4] With OneAgent 1.301 or later, pods on Full-Stack monitored hosts are included in Full-Stack Monitoring at no additional charge. Pods on other hosts, and pods stuck in Pending, still consume pod-hours.[4] Kubernetes warning events are included under fair use, about 60 events per pod-hour pooled across all pods. Dynatrace may charge for events if usage materially exceeds fair use.[4] At USD 0.002 per pod-hour, 500 pods running continuously cost USD 720 per month at list price.[4] Catalog: Kubernetes Platform Monitoring is included for pods on Full-Stack hosts; Pod-hours do not depend on pod size or container count.
Code Monitoring, which is mainly used through Live Debugger, is charged per container-hour. On hosts without containers, each concurrent instrumented process counts separately.[5]
Logs
Log ingest is measured on the raw bytes sent to Dynatrace before enrichment and transformation, per GiB.[7] Customers can choose between two pricing models. On the usage-based model, retention costs USD 0.0007 per GiB-day and queries USD 0.0035 per GiB scanned. On the Retain with Included Queries model, retention costs USD 0.02 per GiB-day with queries included.[6] Query cost depends on the uncompressed data read. The highest possible cost of a query is the log volume in its search range times the rate-card price.[9] Retain with Included Queries covers only the first 10 to 35 days of the configured retention. Storage beyond that is billed at the usage-based retain rate.[8] Catalog: Log ingest is measured on raw bytes before enrichment; Retain with Included Queries covers queries only for the first 10 to 35 days.
Real User Monitoring and application security
Real User Monitoring charges per session. A visit that spans several applications counts one session per application, and a visit longer than an hour counts an extra session for each started hour. Bounced single-action visits do not count.[10] Twenty session properties per application are included, and additional properties are billed per session in which they appear.[10] Runtime Vulnerability Analytics uses memory-GiB-hours with the same 0.25 GiB rounding and 4 GiB host minimum as Full-Stack Monitoring.[11] Catalog: RUM sessions are counted per application, and bounced sessions are free.
Floors. The only floors are the 4 GiB host minimum, the 256 MiB container minimum and the 15-minute interval.[1]
Virtualization & partitioning
Virtual machines are measured like physical hosts, by their virtual RAM.[1] No hypervisor, socket or core count is used. A large VM consolidating many workloads is therefore charged on its memory under Full-Stack and as one host under Infrastructure Monitoring. For Kubernetes, the unit depends on where the pod runs: included on Full-Stack hosts, charged per pod-hour elsewhere.[4] See virtualization and partitioning.
Cloud / BYOL
Managed container services are measured per container where Dynatrace detects them. Azure App Service on Linux with OneAgent 1.297+ and automatic container detection uses the container’s used memory. Without detection, it uses the plan instance’s memory multiplied by the number of running containers.[1] Bring-your-own-licence does not apply. The commitment model is covered in Dynatrace Platform Subscription and rate card.
Programs
All the capabilities above draw down from the DPS annual commitment and continue on demand at the same rates once it is consumed.[12]
Out of scope
- Synthetic monitoring, automation workflows, AppEngine functions, data egress, Security Posture Management and Mainframe Monitoring, apart from their rate-card units.
- Metrics, traces and events powered by Grail as separate capabilities, beyond the included volumes described above.
- Classic host units, which are covered in Dynatrace classic licensing: host units, DDUs and DEM units.