R Server (Standalone)
In SQL Server 2016, Microsoft released R Server (Standalone), as part of its platform for supporting enterprise-class analytics. Microsoft R Server provides scalability and security for the R language, and addresses the in-memory limitations of open source R. Like SQL Server R Services, Microsoft R Server (Standalone) provides parallel and chunked processing of data, enabling R users to use data much bigger than can fit in memory.
In SQL Server 2017, support has been added for the Python language, which enjoys broad support in the machine community learning, and includes popular libraries for text analytics and deep learning. To reflect this broader set of languages, we've also renamed it to Microsoft Machine Learning Server (Standalone).
Benefits of Microsoft R Server
You can use Microsoft R Server for distributed computing on multiple platforms. When you install from SQL Server setup, you get the Windows-based server and all the required tools for publishing and deploying models. For more information about other platforms, see these resources in the MSDN library:
You can also install Microsoft R Server to use as a development client, to get the RevoScaleR libraries and tools needed to create R solutions that can be deployed to SQL Server.
What's new in Microsoft Machine Learning Server
If you install Machine Learning Services (Standalone) using SQL Server 2017 setup, you now have the option to add the Python language. Naturally, the R language is still a supported option, and you can even install both languages if desired.
In SQL Server 2017 CTP 2.0, the server installation also includes the mrsdeploy package and other utilities used for operationalizing models. For more information, see Operationalization with mrsdeploy.
Enterprise users of SQL Server Machine Learning can use the downloadable installers for Microsoft R Server to upgrade their R components, in a process called binding. For more information, see Use SqlBindR to Upgrade and Instance of SQL Server
Get Microsoft R Server or Machine Learning Server (Standalone)
There are multiple options for installing Microsoft R Server:
Use the SQL Server setup wizard
Run SQL Server 2016 setup to install Microsoft R Server (Standalone). The R language is added by default. Or, run SQL Server 2017 setup to install Machine Learning Server (Standalone) and select either R or Python, or both.
The option to install the server is in the Shared Features section of setup. Do not install any other components.
Preferably, do not install the server on a computer where SQL Server R Services or SQL Server Machine Learning Services has been installed.
Use command-line options for SQL Server setup
SQL Server setup supports unattended installs via a rich set of command-line arguments.
Use the standalone installer
You can now use a new Windows installer to set up a new instance of Microsoft R Server or Microsoft Machine Learning Server. Microsoft R Server (and Microsoft Machine Learning Server) requires a SQL Server Enterprise agreement. However, after you run the standalone installer, the support policy for an existing installation is updated, to use the new Modern Lifecycle policy. This support option ensures that updates to machine learning components are applied more frequently than they would be when using the SQL Server service releases.
Upgrade a SQL Server instance
You can use the standalone installer to upgrade an instance of SQL Server 2016 R Services to use the latest version of R. When you run the installer, the Modern Lifecycle support policy will be applied to the server, and the R components will get more frequent updates.
[!NOTE} Currently this update method is available only for existing installations of SQL Server 2016. However, upgrades will be supported for SQL Server 2017 in future.
Related machine learning products
Azure virtual machines with R Server
The Azure marketplace includes multiple virtual machine images that include R Server. Creating a new R Server virtual machine in Microsoft Azure is the fastest way to set up a server to use in developing and deploying predictive models. Images come with features for scaling and sharing already configured, which makes it easier to embed R analytics inside applications and to integrate R with backend systems.
Data Science Virtual Machine
The latest version of the Data Science Virtual machine includes R Server, SQL Server, plus an array of the most popular tools for machine learning, all preinstalled and tested. Create Jupyter notebooks, develop solutions in Julia, and use GPU-enabled deep learning libraries like MXNet, CNTK, and TensorFlow.
For samples, tutorials, and more information about Microsoft R Server, see Microsoft R Products.