SQL Server 2016 contains important improvements to the In-Memory OLTP engine. I will mention the new features that in my opinion will have a great impact on the adoption of In-Memory OLTP.
Max 2TB of durable in-memory data
In-Memory OLTP implementations which need to store more than 256GB (in SQL Server 2014) of business-critical data can host 2TB of durable data in memory-optimized tables.
Support for LOB data types
The maximum row size in a memory-optimized table is 8KB (8060 bytes) in SQL Server 2014. By supporting LOB data types customers can store more data in memory-optimized tables without implementing workarounds for their wide columns.
Greater T-SQL surface area
The migration from traditional RDBMS implementations to In-Memory OLTP becomes easier. SQL Server 2016 supports more T-SQL in in-memory database objects. Foreign keys, outer joins, subqueries, check | unique constraints, natively-compiled user defined functions, natively-compiled after triggers and other T-SQL features are supported.
Multiple offline checkpoint threads
The asynchronous checkpoint process which is responsible for writing in-memory log records to the checkpoint files can’t keep up with high speed synchronous writes to the transaction log. The throughput of the offline checkpoint process is improved by supporting a checkpoint thread for each container of the in-memory filegroup.
Database professionals will be able to alter and recompile natively-compiled stored procedures and memory-optimized tables. In case of memory-optimized tables there will be partial support for the offline alter capabilities.
In-Memory OLTP matures! Great news for database professionals 🙂