Public Preview: Performance monitoring for Azure SQL

Today we’re excited to announce the public preview of performance monitoring for Azure SQL.

Performance monitoring gives you deep visibility into the health and performance of your SQL estate, built directly into the Azure SQL platform. It works across:

• Azure SQL Database


• Azure SQL Managed Instance (coming soon)


• SQL Server on Azure Virtual Machines


• SQL Server enabled by Azure Arc

Microsoft collects performance-related telemetry, runs the pipeline, and stores the data for you. There’s nothing you need to deploy and nothing to operate. You get direct query access to your telemetry. And in Fabric Database Hub, you get prebuilt dashboards and a single view of your whole estate.

 

 

Why we built this

Monitoring SQL performance at scale often meant building and running your own monitoring stack. Before you could effectively answer, “which of my databases need attention right now?”, you typically had to:

1. Deploy and configure a collection resource, such as a watcher or an agent


2. Build a telemetry pipeline to move the data


3. Provision a data store, and then pay for it, secure it, and keep it running


4. Build dashboards on top of all of it


5. Repeat for every new server, database, or region

That’s a lot of work before you see your first chart. And every step is another thing that can break, drift, or quietly stop collecting data.

Customers are also turning to AI to make sense of their database estate. They want an AI agent that can spot a performance problem, explain what’s causing it, and recommend a fix. But an AI agent is only as good as the data it can reach. It operates best with one consistent source of performance data across every database, not a patchwork of tools and data stores.

We heard this feedback loud and clear from customers. You told us you love having at-scale dashboards, query-level visibility, and ownership of your performance data. You also told us that setting up a telemetry stack was time consuming, scale limits got in the way, and running the data store added operational burden and cost overhead. One customer put it simply: They wanted to spend less time managing their telemetry infrastructure and more time managing and improving their databases.

Performance monitoring keeps the parts you valued and removes the infrastructure you had to manage.

No infrastructure to manage

With performance monitoring, you don’t need to:

• Create or manage watcher resources or collection agents


• Build or operate telemetry pipelines


• Provision, size, or pay for your own data store


• Worry about scale limits on how many targets you can monitor

It’s all managed by Microsoft.

Telemetry is collected close to the database engine and sent to a Microsoft-managed telemetry pipeline and data store. Access to that data is governed by Azure role-based access control (RBAC), so people only see telemetry for the resources they already have access to.

Key capabilities

Consistent telemetry across your SQL estate

Performance monitoring collects the same core set of performance data across every supported SQL deployment, whether it runs in Azure, on-premises, or in another cloud. That means one mental model and one set of dashboards, instead of a different tool for every flavor of SQL.

The preview collects performance-related telemetry including:

• CPU and memory utilization


• Wait statistics


• Active sessions


• Storage I/O and database storage utilization


• Performance counters


• Client connections


• Database properties


• Availability group, replica, and database replica health

Prebuilt dashboards in Fabric Database Hub

Performance monitoring comes with prebuilt dashboards in Fabric Database Hub, so you can see performance at a glance without building anything yourself. The dashboards are designed to quickly answer two questions:

1. Are my databases healthy?


2. Which ones need my attention?

From there, you can drill into resource usage, waits, and session activity to understand what’s driving a change in performance.

Query your telemetry directly

Your telemetry is available through a Microsoft-managed, RBAC-governed Azure Data Explorer endpoint. You can connect from the Azure Data Explorer web UI and query it with Kusto Query Language (KQL). You don’t need to create or pay for your own Azure Data Explorer cluster.

This opens a lot of options:

• Build your own reports and dashboards


• Connect tools you already use, such as Grafana or Power BI


• Give an AI agent access to investigate performance across your estate

Here’s a quick example that ranks the resources you can see by 95th-percentile CPU over the last hour:

SqlServerCPUUtilization
| where SampleTimeUTC > ago(1h)
| summarize
AvgCPU = round(avg(AvgCPUPercent), 1)
, P95CPU = round(percentile(AvgCPUPercent, 95), 1)
by ResourceID, ResourceTypeK
| top 10 by P95CPU desc

To get started, see Query performance monitoring telemetry. The article includes the schema, connection steps, and a set of ready-to-run starter queries.

 

 

See your entire database estate in Fabric Database Hub

Performance monitoring is integrated with Fabric Database Hub, where you can see your entire database estate from Azure in one place, including:

• Azure SQL


• Azure Database for PostgreSQL


• Azure Cosmos DB

Fabric Database Hub is built for when you need to look across all your databases, not just one at a time. With Fabric Database Hub, you can:

• Use prebuilt performance dashboards for your SQL resources


• Get estate-wide visibility into health and performance across database types


• Investigate root cause and query performance across many databases


• Use AI-assisted analysis to find and explain issues faster


• Build Real-Time Dashboards on top of your performance telemetry

Performance monitoring data flows into Fabric Database Hub automatically. There’s no separate onboarding step to connect the two. To learn more about Fabric Database Hub, click here.

Getting started

How you turn on performance monitoring depends on the resource type.

 

To view performance monitoring data, you need:

• The Reader role, or a role with higher privileges, on each subscription that contains the resources you want to query.


• The Microsoft.AzureArcData resource provider registered on the subscription.

Availability

Performance monitoring is available in public preview in the following Azure regions:

• Americas: Brazil South, Canada Central, Canada East, Central US, East US, East US 2, North Central US, South Central US, West Central US, West US, West US 2, West US 3


• Europe, Middle East, and Africa: France Central, North Europe, Norway East, South Africa North, Sweden Central, Switzerland North, UAE North, UK South, UK West, West Europe


• Asia Pacific: Australia East, Central India, Japan East, Korea Central, Southeast Asia

Learn more

• Query performance monitoring telemetry


• Fabric Database Hub


• Monitor SQL Server enabled by Azure Arc


• Supplemental Terms of Use for Microsoft Azure Previews