Deploying Machine Learning Services

SQL Server 2017 Machine Learning Services deployments can be placed in any part of a database application workflow. Machine Learning Services is a built-in analytic component that provides advanced analytic features, including predictive modeling and machine learning capabilities through the use of Python and R language scripts. The illustration shows two typical uses of Machine Learning in a SQL Server production deployment.

In configuration 1 (top), incoming data from credit card transactions are received by SQL Server and analyzed with a Machine Learning Service fraud detection script, prior to saving the data and results in the production database. This type of configuration allows developers to include machine learning scripts as part of normal data-processing routines, without passing the data to a separate server for analysis.

In configuration 2 (bottom), SQL Server is deployed with an AlwaysOn Availability Group (AG), which automatically replicates data to secondary servers. In this example, real-time data is replicated from the Primary to Replica 2 for Reporting Services workloads and to Replica 3 for Machine Learning Services workloads. This type of configuration ensures that performance on the primary database on Replica 1 is not impacted by the other workloads.