Install SQL Server Machine Learning Services on Windows
This article explains how to install SQL Server Machine Learning Services on Windows. You can use Machine Learning Services to execute Python and R scripts in-database.
SQL Server 2017 (or greater) Setup is required if you want to install Machine Learning Services with R or Python language support. If instead you have SQL Server 2016 installation media, you can install SQL Server R Services (In-Database) to get R language support.
A database engine instance is required. You cannot install just R or Python features, although you can add them incrementally to an existing instance.
For business continuity, Always On Availability Groups are supported for Machine Learning Services. You have to install Machine Learning Services, and configure packages, on each node.
Installing Machine Learning Services is not supported on a failover cluster in SQL Server 2017. However, it is supported with SQL Server 2019.
Do not install Machine Learning Services on a domain controller. The Machine Learning Services portion of setup will fail.
Do not install Shared Features > Machine Learning Server (Standalone) on the same computer running an in-database instance. A standalone server will compete for the same resources, undermining the performance of both installations.
Side-by-side installation with other versions of R and Python is supported but not recommended. It's supported because SQL Server instance uses its own copies of the open-source R and Anaconda distributions. But it's not recommended because running code that uses R and Python on the SQL Server computer outside SQL Server can lead to various problems:
- You use a different library and different executable, and get different results, than you do when you are running in SQL Server.
- R and Python scripts running in external libraries cannot be managed by SQL Server, leading to resource contention.
After setup is complete, be sure to complete the post-configuration steps described in this article. These steps include enabling SQL Server to use external scripts, and adding accounts required for SQL Server to run R and Python jobs on your behalf. Configuration changes generally require a restart of the instance, or a restart of the Launchpad service.
Get the installation media
The download location for SQL Server depends on the edition:
- SQL Server Enterprise, Standard, and Express Editions are licensed for production use. For the Enterprise and Standard Editions, contact your software vendor for the installation media. You can find purchasing information and a directory of Microsoft partners on the Microsoft purchasing website.
- Free editions are available at SQL Server downloads.
For local installations, you must run Setup as an administrator. If you install SQL Server from a remote share, you must use a domain account that has read and execute permissions on the remote share.
Start the setup wizard for SQL Server 2017.
On the Installation tab, select New SQL Server stand-alone installation or add features to an existing installation.
On the Feature Selection page, select these options:
Database Engine Services
To use R and Python with SQL Server, you must install an instance of the database engine. You can use either a default or a named instance.
Machine Learning Services (In-Database)
This option installs the database services that support R and Python script execution.
Check this option to add the Microsoft R packages, interpreter, and open-source R.
Check this option to add the Microsoft Python packages, the Python 3.5 executable, and select libraries from the Anaconda distribution.
Do not select the option for Machine Learning Server (Standalone). The option to install Machine Learning Server under Shared Features is intended for use on a separate computer.
On the Consent to Install R page, select Accept. This license agreement covers Microsoft R Open, which includes a distribution of the open-source R base packages and tools, together with enhanced R packages and connectivity providers from the Microsoft development team.
On the Consent to Install Python page, select Accept. The Python open-source licensing agreement also covers Anaconda and related tools, plus some new Python libraries from the Microsoft development team.
If the computer you are using does not have internet access, you can pause setup at this point to download the installers separately. For more information, see Install machine learning components without internet access.
Select Accept, wait until the Next button becomes active, and then select Next.
On the Ready to Install page, verify that these selections are included, and select Install.
- Database Engine Services
- Machine Learning Services (In-Database)
- R or Python, or both
Note of the location of the folder under the path
..\Setup Bootstrap\Logwhere the configuration files are stored. When setup is complete, you can review the installed components in the Summary file.
After setup is complete, if you are instructed to restart the computer, do so now. It is important to read the message from the Installation Wizard when you have finished with Setup. For more information, see View and Read SQL Server Setup Log Files.
Set environment variables
For R feature integration only, you should set the MKL_CBWR environment variable to ensure consistent output from Intel Math Kernel Library (MKL) calculations.
In Control Panel, click System and Security > System > Advanced System Settings > Environment Variables.
Create a new User or System variable.
- Set variable name to
- Set the variable value to
This step requires a server restart. If you are about to enable script execution, you can hold off on the restart until all of the configuration work is done.
Enable script execution
Open SQL Server Management Studio.
Connect to the instance where you installed Machine Learning Services, click New Query to open a query window, and run the following command:
The value for the property,
external scripts enabled, should be 0 at this point. That is because the feature is turned off by default. The feature must be explicitly enabled by an administrator before you can run R or Python scripts.
To enable the external scripting feature, run the following statement:
EXEC sp_configure 'external scripts enabled', 1 RECONFIGURE WITH OVERRIDE
If you have already enabled the feature for the R language, don't run reconfigure a second time for Python. The underlying extensibility platform supports both languages.
Restart the service
When the installation is complete, restart the database engine before continuing to the next, enabling script execution.
Restarting the service also automatically restarts the related SQL Server Launchpad service.
You can restart the service using the right-click Restart command for the instance in SSMS, or by using the Services panel in Control Panel, or by using SQL Server Configuration Manager.
Check the installation status of the instance in custom reports or setup logs.
Use the following steps to verify that all components used to launch external script are running.
In SQL Server Management Studio, open a new query window, and run the following command:
EXEC sp_configure 'external scripts enabled'
The run_value should now be set to 1.
Open the Services panel or SQL Server Configuration Manager, and verify SQL Server Launchpad service is running. You should have one service for every database engine instance that has R or Python installed. For more information about the service, see Extensibility framework.
If Launchpad is running, you should be able to run simple R and Python scripts to verify that external scripting runtimes can communicate with SQL Server.
Open a new Query window in SQL Server Management Studio, and then run a script such as the following:
- For R
EXEC sp_execute_external_script @language =N'R', @script=N' OutputDataSet <- InputDataSet; ', @input_data_1 =N'SELECT 1 AS hello' WITH RESULT SETS (([hello] int not null)); GO
- For Python
EXEC sp_execute_external_script @language =N'Python', @script=N' OutputDataSet = InputDataSet; ', @input_data_1 =N'SELECT 1 AS hello' WITH RESULT SETS (([hello] int not null)); GO
The script can take a little while to run, the first time the external script runtime is loaded. The results should be something like this:
Columns or headings used in the Python script are not returned, by design. To add column names for your output, you must specify the schema for the return data set. Do this by using the WITH RESULTS parameter of the stored procedure, naming the columns and specifying the SQL data type.
For example, you can add the following line to generate an arbitrary column name:
WITH RESULT SETS ((Col1 AS int))
We recommend that you apply the latest cumulative update to both the database engine and machine learning components.
On internet-connected devices, cumulative updates are typically applied through Windows Update, but you can also use the steps below for controlled updates. When you apply the update for the database engine, Setup pulls cumulative updates for any R or Python features you installed on the same instance.
On disconnected servers, extra steps are required. For more information, see Install on computers with no internet access > Apply cumulative updates.
Start with a baseline instance already installed: SQL Server 2017 initial release
Go to the cumulative update list: SQL Server 2017 updates
Select the latest cumulative update. An executable is downloaded and extracted automatically.
Run Setup. Accept the licensing terms, and on the Feature selection page, review the features for which cumulative updates are applied. You should see every feature installed for the current instance, including machine learning features. Setup downloads the CAB files necessary to update all features.
- Continue through the wizard, accepting the licensing terms for R and Python distributions.
If the external script verification step was successful, you can run R or Python commands from SQL Server Management Studio, Visual Studio Code, or any other client that can send T-SQL statements to the server.
If you got an error when running the command, review the additional configuration steps in this section. You might need to make additional appropriate configurations to the service or database.
At the instance level, additional configuration might include:
- Firewall configuration for SQL Server Machine Learning Services
- Enable additional network protocols
- Enable remote connections
- Create a login for SQLRUserGroup
- Manage disk quotas to avoid external scripts running tasks that exhaust disk space
In SQL Server 2019 on Windows, the isolation mechanism has changed. This affects SQLRUserGroup, firewall rules, file permission, and implied authentication. For more information, see Isolation changes for Machine Learning Services.
On the database, you might need the following configuration updates:
Whether additional configuration is required depends on your security schema, where you installed SQL Server, and how you expect users to connect to the database and run external scripts.
Now that you have everything working, you might also want to optimize the server to support machine learning, or install pretrained models.
Add more worker accounts
If you expect many users to be running scripts concurrently, you can increase the number of worker accounts that are assigned to the Launchpad service. For more information, see Modify the user account pool for SQL Server Machine Learning Services.
Optimize the server for script execution
The default settings for SQL Server setup are intended to optimize the balance of the server for a variety of services that are supported by the database engine, which might include extract, transform, and load (ETL) processes, reporting, auditing, and applications that use SQL Server data. Therefore, under the default settings, you might find that resources for machine learning are sometimes restricted or throttled, particularly in memory-intensive operations.
To ensure that machine learning jobs are prioritized and resourced appropriately, we recommend that you use SQL Server Resource Governor to configure an external resource pool. You might also want to change the amount of memory that's allocated to the SQL Server database engine, or increase the number of accounts that run under the SQL Server Launchpad service.
To configure a resource pool for managing external resources, see Create an external resource pool.
To change the amount of memory reserved for the database, see Server memory configuration options.
To change the number of R accounts that can be started by SQL Server Launchpad, see Modify the user account pool for machine learning.
If you are using Standard Edition and do not have Resource Governor, you can use Dynamic Management Views (DMVs) and Extended Events, as well as Windows event monitoring, to help manage the server resources. For more information, see Monitoring and managing R Services and Monitoring and managing Python Services.
Install additional R packages
The R solutions you create for SQL Server can call basic R functions, functions from the proprietary packages installed with SQL Server, and third-party R packages compatible with the version of open-source R installed by SQL Server.
Packages that you want to use from SQL Server must be installed in the default library that is used by the instance. If you have a separate installation of R on the computer, or if you installed packages to user libraries, you won't be able to use those packages from T-SQL.
To install and manage R packages, you can set up user groups to share packages on a per-database level, or configure database roles to enable users to install their own packages. For more information, see Install new R packages in SQL Server.
R developers can get started with some simple examples, and learn the basics of how R works with SQL Server. For your next step, see the following links:
Python developers can learn how to use Python with SQL Server by following these tutorials:
To view examples of machine learning that are based on real-world scenarios, see Machine learning tutorials.