In-Memory OLTP architecture

The architecture of  In-Memory OLTP on SQL Server is completely different from everything seen before on SQL Server. Authors of the solution proceeded from the assumption that everything that could be done on the disk-based part of the data engine was just an evolution of the data processing optimization, so some kind of revolution was needed. In this chapter, we will explain what In-Memory OLTP is, we will describe the top-level architecture, and then we'll jump deeper into the details of In-Memory OLTP.

To see the benefits gained when using In-Memory OLTP, we need to consider the differences between two areas of work: data storage architecture and user request processing phases.

Data storage differences in OLTP

Disk-based data is stored in classic files ( .mdf or .ndf) and divided into small 8 KB parts called data pages. Those data pages are registered within other data pages called Index Allocation Maps (IAMs), and other types of data pages are used for identification...