Azure Monitor Metrics overview

Azure Monitor Metrics is a feature of Azure Monitor that collects numeric data from monitored resources into a time-series database. Metrics are numerical values that are collected at regular intervals and describe some aspect of a system at a particular time.

Types of metrics

There are multiple types of metrics supported by Azure Monitor Metrics:

• Native metrics use tools in Azure Monitor for analysis and alerting.
  • Platform metrics are collected from Azure resources. They require no configuration and have no cost.
• Prometheus metrics are collected from Kubernetes clusters including Azure Kubernetes service (AKS) and use industry standard tools for analyzing and alerting such as PromQL and Grafana.

The differences between each of the metrics are summarized in the following table.

Category	Native platform metrics	Prometheus metrics
Sources	Azure resources	Azure Kubernetes service (AKS) cluster Any Kubernetes cluster through remote-write
Configuration	None	Enable Azure Monitor managed service for Prometheus
Stored	Subscription	Azure Monitor workspace
Cost	No	Yes (free during preview)
Aggregation	preaggregated	raw data
Analyze	Metrics Explorer	PromQL Grafana dashboards
Alert	metrics alert rule	Prometheus alert rule
Visualize	Workbooks Azure dashboards	Grafana
Retrieve	Azure CLI Azure PowerShell cmdlets REST API or client library .NET Go Java JavaScript Python	Grafana

Data collection

Azure Monitor collects metrics from the following sources. After these metrics are collected in the Azure Monitor metric database, they can be evaluated together regardless of their source:

• Azure resources: Platform metrics are created by Azure resources and give you visibility into their health and performance. Each type of resource creates a distinct set of metrics without any configuration required. Platform metrics are collected from Azure resources at one-minute frequency unless specified otherwise in the metric's definition.
• Applications: Application Insights creates metrics for your monitored applications to help you detect performance issues and track trends in how your application is being used. Values include Server response time and Browser exceptions.
• Virtual machine agents: Metrics are collected from the guest operating system of a virtual machine. You can enable guest OS metrics for Windows virtual machines by using the Azure Monitor Agent. Azure Monitor Agent replaces the legacy agents - Windows diagnostic extension and the InfluxData Telegraf agent for Linux virtual machines.
• Kubernetes clusters: Kubernetes clusters typically send metric data to a local Prometheus server that you must maintain. Azure Monitor managed service for Prometheus provides a managed service that collects metrics from Kubernetes clusters and store them in Azure Monitor Metrics.

REST API

Azure Monitor provides REST APIs that allow you to get data in and out of Azure Monitor Metrics.

• Azure Monitor Metrics REST API - Allows you to access Azure Monitor platform metrics definitions and values. For more information, see Azure Monitor REST API. For information on how to use the API, see the Azure monitoring REST API walkthrough.
• Azure Monitor Metrics Batch REST API - Azure Monitor Metrics Batch API is a high-volume API designed for customers with large volume metrics queries. It's similar to the existing standard Azure Monitor Metrics REST API, but provides the capability to retrieve metric data for up to 50 resource IDs in the same subscription and region in a single batch API call. This improves query throughput and reduces the risk of throttling.

Security

All communication between connected systems and the Azure Monitor service is encrypted using the TLS 1.2 (HTTPS) protocol. The Microsoft SDL process is followed to ensure all Azure services are up-to-date with the most recent advances in cryptographic protocols.

Secure connection is established between the agent and the Azure Monitor service using certificate-based authentication and TLS with port 443. Azure Monitor uses a secret store to generate and maintain keys. Private keys are rotated every 90 days and are stored in Azure and are managed by the Azure operations who follow strict regulatory and compliance practices. For more information on security, see Encryption of data in transit, Encryption of data at rest, and Azure Monitor Logs data security

Metrics Explorer

Use Metrics Explorer to interactively analyze the data in your metric database and chart the values of multiple metrics over time. You can pin the charts to a dashboard to view them with other visualizations. You can also retrieve metrics by using the Azure monitoring REST API.

For more information, see Analyze metrics with Azure Monitor metrics explorer.

Data structure

Data that Azure Monitor Metrics collects, is stored in a time-series database that's optimized for analyzing time-stamped data. Each set of metric values is a time series with the following properties:

• The time when the value was collected.
• The resource that the value is associated with.
• A namespace that acts like a category for the metric.
• A metric name.
• The value itself.
• Multiple dimensions when they're present.

Multi-dimensional metrics

One of the challenges to metric data is that it often has limited information to provide context for collected values. Azure Monitor addresses this challenge with multi-dimensional metrics.

Metric dimensions are name/value pairs that carry more data to describe the metric value. For example, a metric called Available disk space might have a dimension called Drive with values C: and D:. That dimension would allow viewing available disk space across all drives or for each drive individually.

See Apply dimension filters and splitting for details on viewing metric dimensions in metrics explorer.

Nondimensional metric

The following table shows sample data from a nondimensional metric, network throughput. It can only answer a basic question like "What was my network throughput at a given time?"

Timestamp	Metric value
8/9/2017 8:14	1,331.8 Kbps
8/9/2017 8:15	1,141.4 Kbps
8/9/2017 8:16	1,110.2 Kbps

Network throughput and two dimensions ("IP" and "Direction")

The following table shows sample data from a multidimensional metric, network throughput with two dimensions called IP and Direction. It can answer questions such as "What was the network throughput for each IP address?" and "How much data was sent versus received?"

Timestamp	Dimension "IP"	Dimension "Direction"	Metric value
8/9/2017 8:14	IP="192.168.5.2"	Direction="Send"	646.5 Kbps
8/9/2017 8:14	IP="192.168.5.2"	Direction="Receive"	420.1 Kbps
8/9/2017 8:14	IP="10.24.2.15"	Direction="Send"	150.0 Kbps
8/9/2017 8:14	IP="10.24.2.15"	Direction="Receive"	115.2 Kbps
8/9/2017 8:15	IP="192.168.5.2"	Direction="Send"	515.2 Kbps
8/9/2017 8:15	IP="192.168.5.2"	Direction="Receive"	371.1 Kbps
8/9/2017 8:15	IP="10.24.2.15"	Direction="Send"	155.0 Kbps
8/9/2017 8:15	IP="10.24.2.15"	Direction="Receive"	100.1 Kbps

Retention of metrics

Platform metrics

Platform metrics are stored for 93 days with the following exceptions:

• Classic guest OS metrics: These performance counters are collected by the Windows diagnostic extension or the Linux diagnostic extension and routed to an Azure Storage account. Retention for these metrics is guaranteed to be at least 14 days, although no expiration date is written to the storage account.

  For performance reasons, the portal limits how much data it displays based on volume. So, the actual number of days that the portal retrieves can be longer than 14 days if the volume of data being written isn't large.

• Guest OS metrics sent to Azure Monitor Metrics: These performance counters are collected by the Windows diagnostic extension and sent to the Azure Monitor data sink, or the InfluxData Telegraf agent on Linux machines, or the newer Azure Monitor agent via data-collection rules. Retention for these metrics is 93 days.

• Guest OS metrics collected by the Log Analytics agent: These performance counters are collected by the Log Analytics agent and sent to a Log Analytics workspace. Retention for these metrics is 31 days and can be extended up to 2 years.

• Application Insights log-based metrics: Behind the scenes, log-based metrics translate into log queries. Their retention is variable and matches the retention of events in underlying logs, which is 31 days to 2 years. For Application Insights resources, logs are stored for 90 days.

While platform metrics are stored for 93 days, you can only query (in the Metrics tile) for a maximum of 30 days' worth of data on any single chart. This limitation doesn't apply to log-based metrics. If you see a blank chart or your chart displays only part of metric data, verify that the difference between start and end dates in the time picker doesn't exceed the 30-day interval. After you've selected a 30-day interval, you can pan the chart to view the full retention window.

Prometheus metrics

Prometheus metrics are stored for 18 months, but a PromQL query can only span a maximum of 32 days.

Next steps

• Learn more about the Azure Monitor data platform.
• Learn about log data in Azure Monitor.
• Learn about the monitoring data available for various resources in Azure.