DDUs for metrics

This page explains how we calculate DDUs for metrics, the concept of custom metrics, and how you can estimate and track DDU consumption for metrics.

Which types of metrics consume DDUs?

• Custom metrics
• Certain built-in metrics (see below)

What is a custom metric?

Custom metrics are metrics that you define or install. They extend the value of Dynatrace monitoring capabilities by enabling you to integrate third-party data sources, calculate custom metrics, import metrics via API, leverage extensions, and more.

Following is a non-exhaustive list of custom metric types:

• All API-ingested metrics
• Calculated service metrics, custom DEM metrics, and log metrics
  • For example, calculated service metrics created upon request attributes
• All metrics provided by manually installed Dynatrace Monitoring Extensions

For full details on the setup and ingestion of custom metrics in Dynatrace, see Metric ingestion.

Which built-in metrics consume DDUs?

Metrics that consume DDUs include:

• Metrics provided by certain built-in extensions that are not enabled by default
  • Dynatrace web UI switches to enable/disable these metrics are labeled to let you know if they affect DDU consumption.
• All metrics ingested by the Public Cloud Extensions (AWS, Azure and GCP).
  • For example, metrics derived from cloud services.

The built-in metrics page lists all metrics that do not consume DDUs.

How do we calculate DDU consumption for metrics?

Consumption occurs when metric data points are ingested. A metric data point is a single measurement value stored in Dynatrace. The value belongs to a metric, which is identified by a metric key and has a timestamp. A time series metric is a series of such data points, for example CPU utilization for a given host across an analysis timeframe.

Each ingested data point deducts 0.001 DDU from your available quota.

Metrics provided by Dynatrace are typically written once per minute. This results in the following typical annual DDU consumption for a single metric:

1 metric data point x 60 min x 24 h x 365 days x 0.001 = 525.6 DDUs per metric/year

How do dimensions of a metric impact the DDU consumption?

Metric data points can have associated dimensions, which are identified by dimension keys and values. For example, using the syntax of the metrics ingestion protocol two values could be ingested as follows:

plaintextcopydownload
my_cpu_utilization,hostname=hostA 55 1609459200000
my_cpu_utilization,hostname=hostB 45 1609459200000

For the same timestamp (1609459200000), Dynatrace retains two different measurement values for the same metric on two different hosts. This is achieved using the dimension key hostname and dimension values hostA and hostB. 55 and 45 are the measured CPU utilization percentages for this metric on the two hosts.

Notice how you can consider this as two metrics—one chart can be created for the CPU utilization of hostA and another for the CPU utilization of hostB. If data points are ingested periodically for both hosts, the end effect is double the number of ingested data points—because here each host-dimension results in a separate metric value that must be retained.

This doubling of ingested data points is also reflected in the associated DDU consumption: The DDU consumption is now effectively 0.002 per minute because data from two hosts is being ingested. Therefore, to calculate the DDU consumption based on metric keys, you need consider which data is being ingested and take the data's dimensions into account.

If more dimensions are added, and the number of data points increases, then DDU consumption will also increase. Here's an example:

plaintextcopydownload
my_cpu_utilization,hostname=hostA,cpu=1 55 1609459200000
my_cpu_utilization,hostname=hostA,cpu=2 11 1609459200000
my_cpu_utilization,hostname=hostB,cpu=1 45 1609459200000
my_cpu_utilization,hostname=hostB,cpu=2 45 1609459200000

This example will consume 4 DDUs per minute due to the two hosts and the two CPUs.

However, the additional consumption is not caused by the increase in dimensions, but rather by the increased number of ingested data points, which typically accompany an increased number of dimensions.

If dimensions are added, but the number of data points remains the same, then DDU consumption does not change. This happens when purely descriptive dimensions are added whose values are determined by other pre-existing dimensions:

plaintextcopydownload
my_cpu_utilization,hostname=hostA,host_ip="127.0.0.23" 55 1609459200000
my_cpu_utilization,hostname=hostB,host_ip="127.0.0.42" 45 1609459200000

These ingested data points amount to the same DDU consumption as if they were sent without the host_ip dimension. Formally speaking, if dimensions are added, whose values are functionally determined by other pre-existing dimensions, then the DDU consumption will not change.

Included metrics per host unit

If you are monitoring hosts with OneAgent and consuming host units, then you are granted a number of "host-included metrics" based on the host unit consumption of the monitored hosts. More precisely, for each given minute, a host consuming host units has a budget of host-included metric data points. When metric data points "about the host" are ingested (for instance with a formal host dimension, then first the host-included budget of data points is consumed before your DDU quota is affected. One host-included metric is defined as one metric data point per minute, considering the previous section about dimensions of ingested data points.

1,000 metrics per host unit are included on each monitored host (in Full-Stack Monitoring mode). These metrics are included with your Full-Stack host units and therefore don't consume DDUs.

• Every OneAgent-monitored host with Full-Stack Monitoring enabled includes 1000 metrics per Host Unit (see detail in the table below).
• OneAgent-monitored hosts with Infrastructure Monitoring enabled always include 200 metrics that do not consume DDUs.
• All metrics that exceed the host's included metrics will consume DDUs.
• If all DDUs in the Metrics pool are used up, then the included metrics per host unit are not available as well.
• Calculated service metrics, custom DEM metrics, and log metrics are not eligible for these included metrics per host unit

Note that included metrics are bound to specific hosts that are monitored via OneAgent. This means that only metrics that are booked either on the monitored host or a process of a monitored host are included. This also includes metrics that are booked onto AWS EC2, Azure VM, and VMware virtual machine entities (if those VMs are monitored by OneAgent). Note that for metrics ingested by API, this applies only to metrics booked on a monitored host.

The evaluation of whether a host exceeds the host-included metrics is based on the total number of metric data points that are reported at a 1-minute frequency (there's no sliding time window for these metrics). When the ingested data points exceed the host-included metrics, DDUs are consumed for the difference.

If a host is licensed based on host-unit hours rather than host units, the calculation works the same way; you can consume up to the number of host-included metrics per host-unit hour. Each data point beyond th this number consumes DDUs.

Example of how included metrics affect DDU consumption

Let's take the following example.

• A host has 16 GB RAM - In other words, 1 host unit
• The host is monitored via the OneAgent in Full-Stack mode

(1 host unit = 1,000 host-included metrics)

• 1,500 metrics from various extensions and local API ingest (for example, JMX and RabbitMQ) are booked on this OneAgent-monitored host
• 300 reported metrics via public API with no relation to a OneAgent-monitored host

In this scenario, 1,000 metrics/min are included via the host unit, and the host reports an excess of 500 metrics which will consume DDUs. The 300 Metrics that are reported via the API, and not related to a OneAgent-monitored host, will also consume DDUs. This makes a total of 800 metrics/min that consume DDUs.

Here are some other simple scenarios (All values below are based on metrics sent at 1 min intervals.)

Reported DDUs vs. consumed DDUs

Reported DDUs are defined as all DDUs before any host-included considerations are performed. Consumed DDUs reflect the total "billable" DDUs that get subtracted from your DDU balance.

You can compare the reported DDUs for a specific host with the number of host-included metrics by creating a chart with the Data explorer using the following metrics. This allows you to gain insights into the number of available vs. consumed DDUs for a specific host.

• builtin:billing.ddu.includedMetricDduPerHost This metric is split by host ID. It shows you the host-included DDUs. For example, if a host has 1,000 included metrics, this metric will have a value of 1 (in other words, 1,000 x 0.001)
• builtin:billing.ddu.metrics.byEntityRaw The byEntityRaw metric is split by monitored entity and reflects the total reported DDUs before any host-included considerations are performed.
• builtin:billing.ddu.metrics.byEntity Similar to the byEntityRaw metric, the byEntity metric is split by monitored entity. It reflects the total consumed DDUs (in other words, all consumption above the number of included metrics per host unit). This is equivalent to the information shown on the DDU consumption overview page (see example below).

Recommendations for working with the chart

• As DDU billing is calculated on a per-minute basis (take a look at the DDU consumption for metrics example for more details) it's recommended that you work with a 1-minute resolution in the chart as well.
• Using Split by and Filter by allows you to select a specific host that you're interested in.

The example Data explorer query above shows the following information:

• The purple line reflects the host-included DDUs. In the above example this is 0.25 DDUs (in other words, 250 metrics) on a per-minute basis.
• The yellow line reflects the byEntityRaw metric. It shows the total reported DDUs for a monitored entity, which is filtered by the host. Any DDU consumption exceeding the host-included budget exceeds the purple line threshold.
• The turquoise line reflects the information from the byEntity metric. It shows you the total consumed metrics. In the example above the host-included limit is exceeded about every 15 minutes between 09:30 and 10:45, and continuously beginning at 10:45. This results in the consumed DDUs, which are represented by the turquoise line chart.

Please note that data points from the byEntity metric can be delayed up to a few minutes.

How to estimate needed DDUs for metrics

Limited metric ingestion and analysis is provided with out-of-the-box Dynatrace technology support in the form of included metrics per host unit. To arrange for additional metric ingestion and analysis, contact Dynatrace Sales.

How to track DDU consumption for metrics

This section introduces two options for tracking the consumption of DDUs for metrics.

Davis data units overview page

To see how many DDUs your environment has consumed

1. In the Dynatrace menu, go to Consumption. Note: You need the Change monitoring settings permission or an admin account to access this page.
2. On the Davis data units (DDU) page:
   • In the Consumption by DDU pool table, check the Metrics row.
   • In the DDU consumption details section, check the Metrics tab.

There you can view your DDU consumption and identify the top contributors to DDU consumption.

Customized dashboards to track consumption

The following use cases explain how to instrument dashboards via Data explorer and provided DDU metrics to track DDU consumption from metrics.

Investigate your total DDU consumption from metrics

Investigate metric consumption per monitored entity

Investigate metric consumption per monitored metric entity

This chart shows you the total of reported DDUs per each metric key.

More information can also be found in the section Reported DDUs vs. consumed DDUs.

Consumption "Not related to a monitored entity"

The source of this consumption is one of the following:

• Metric ingestion for data without a dt.entity.host dimension
• Calculated metrics from user sessions
• Calculated log metrics v2—In the metric-key breakdown, these have a log. prefix

FAQs