Posted by Rob Knies

generic image

As SQL Server 2014 is released to manufacturing on March 18, one of the features that will be bolstering the new release, to be available to customers on April 1, are its in-memory solutions built directly into the product. These solutions began as Hekaton, a collaboration between the product team and Microsoft Research.

One of the selling points for SQL Server 2014 is its fast performance, and the solution built from Hekaton delivers in that regard, with its in-memory online-transaction processing, featuring an updatable in-memory column store and in-memory analytics.

Hekaton is discussed in a video as part of the Microsoft Research Luminaries series on Channel 9. In the video, Paul Larson, Microsoft Research principal researcher, and Mike Zwilling, SQL Server principal architect, chat about the collaboration and the goals for the project.



“There are a lot of things that you can simplify compared with current database systems,” Larson explains. “We removed a lot of the overhead you have in a traditional database system, and we invented a lot of new algorithms to be able to exploit the fact that these are in memory and to be able to scale them to a very, very large number of processes. We need to scale to hundreds of processes.”

Zwilling takes time to discuss what that level of performance enables Hekaton to deliver.

“It excels in random lookups of data,” he says. “It excels in workloads that need very low latency—sub-millisecond measures in hundreds of microseconds—or even less latency to look up rows and to look up data in tables.”

The goals the collaborators had for the Hekaton project were lofty.

"We want to unlock the 10x, 20x, 50x gains for more applications,” Zwilling says. “We're going to go after the wider, broader applications."

For more on their discussion, see the video above. For more background on Hekaton, read this story.