The Microsoft Machine learning R Server platform

We have already touched on the concept of R Open and R for enterprise environment briefly. Microsoft Machine Learning R Server is an enterprise server that delivers high dimensional and large datasets that can be processed in parallel, and the workload can be distributed across nodes. R Server can process these parallel and distributed workloads on Windows, Linux servers, or HDFS systems, such as Hadoop, Spark, Teradata, and HD Insight. R Server can achieve parallel and distributed workloads using Microsoft R packages designed to do just that. The RevoScaleR package will give the ability to do highly parallel and distributed computations, statistical analysis, and predictive analytics, as well as machine learning and deep learning.

With the acquisition of the company Revolution Analytics, Microsoft rebranded their two main products, Revolution R Open and Revolution R Enterprise, to Microsoft R Open and Microsoft R Server and Microsoft SQL Server R Services. In addition to these two flavors, they also added Microsoft R Client as an additional standalone product:

Based on different flavors and enterprise ecosystems, companies can choose the type of usage they will need (community, or non-commercial, and commercial) and, based on their business needs and where they want their analytical server to be set up, they can choose what suits them the most. In the commercial version, a standalone machine learning R Server (or simply R server) is available, as well as in-database machine learning services (or SQL Server R services).

In the version of SQL Server 2017, Microsoft R Server has been changed to Machine Learning Server (both in-database and as a standalone; the rebranding from in-database R Services to Machine Learning R Services was introduced in the CTP version of SQL Server VNext that later changed to SQL Server 2017). In Figure 4, one can see the naming available when installing SQL Server 2016 (left-print screen) and the names that will be available in SQL Server 2017 (right-print screen):

The reason given for the change in the naming was the fact that Microsoft introduced Python in SQL Server 2017. Python is a powerful statistical, analytical, and machine learning language, and therefore, the name was unified. The R-part of the Machine Learning Server will not change internally, but it will gain useful additional packages and improved functions, as follows:

serialize/unserialize -> rxSerializeModel/rxUnserializeModel 
InstallPackages -> rxInstallPackages 
Histogram -> rxHistogram 

All these functions have been rewritten either for working on large datasets in parallel and distributed workloads, or to support R and SQL Server work.

In the Microsoft R platform, the following products are available:

• Microsoft Machine Learning R Server
• Microsoft R Client
• Microsoft R Open
• R Tools for Visual Studio

Product description has been summarized based on Microsoft Docs descriptions and based on an article published on SQLServerCentral in September of 2016 ( http://www.sqlservercentral.com/articles/Microsoft/145393/).