Hekaton : In-Memory Optimization

Hekaton : The codename for a Microsoft project that will provide support for mixing in-memory database tables with more traditional on-disk tables in the same database. These hybrid databases will be able to dynamically handle both in-memory and on-disk storage in the same database for optimal performance and reliability. Traditional RDBMS architecture was designed when memory resources were expensive, and was optimized for disk I/O. Modern hardware has much more memory, which affects database design principles dramatically. Modern design can now optimize for a working set stored entirely in main memory. Hekaton fully provides ACID database properties.

Project Hekaton will enable Microsoft to compete in the in-memory database market with products like Oracle Database's Exadata and Exalytics appliance options and SAP Hana.

This SQL Server In-memory OLTP capability is released in SQL 2014 version. In-Memory OLTP is a memory-optimized OLTP database engine for SQL Server. Depending on the reason for poor performance with your disk-based tables, In-Memory OLTP can help you achieve significant performance and scalability gains by using,

• Algorithms that are optimized for accessing memory-resident data.
• Optimistic concurrency control that eliminates logical locks.
• Lock free objects are used to access all data. Threads that perform transactional work don’t use locks or latches for concurrency control.
• Natively compiled stored procedures result in orders of magnitude reduction in the engine code path.

Use of main memory can result in a few percentage points of performance improvement, to 20 times performance improvement.

Links: http://research.microsoft.com/en-us/news/features/hekaton-122012.aspx, http://msdn.microsoft.com/en-us/library/dn133186%28v=sql.120%29.aspx