Tables and indexes

As explained earlier, Hekaton tables can be accessed either by natively compiled stored procedures or by standard T-SQL, such as ad hoc queries or standard stored procedures. Tables are stored in memory, and each row can potentially have multiple versions. Versions are kept in memory instead of tempdb, which is what the versioning mechanism of the standard database engine uses. Versions that are no longer needed – that is, that are no longer visible to any transaction – are deleted to avoid filling up the available memory. This process is known as garbage collection.

Chapter 5, Working with Indexes, introduced indexes for traditional tables. Memory-optimized tables also benefit from indexes, and in this section, we will talk about these indexes and how they are different from their disk-based counterparts. As explained earlier, Hekaton indexes are never persisted to disk; they only exist in memory, and because of that, their operations are not logged...