How to make SQL Server smoking fast
- By John Breeden II
- Nov 27, 2012
Microsoft SQL Server databases currently rely on traditional hard drives. That’s fine for most applications, but certain ones, such as online transactions, really require a lot of reads and writes. This slows the whole process, adds more wear and tear to disks, and generally causes performance to drag across an entire enterprise. But if all of those transactions could be handled in memory, with no physical layer, this whole process could be considerably speeded up, as much as 50 times faster than it works now.
What it is: Other databases used by public-sector agencies, such as SAP’s HANA and Oracle’s Exadata, allow processes to run in memory, though this mostly affects the big enterprise arena. Small to mid-size operations are almost exclusively run on SQL servers. As such, Microsoft is testing out a new version of SQL, code-named Hekaton, which will let databases run in memory, but only on a single server.
However, there is no upper limit to how much memory can be on a single database server, and administrators can easily designate how much of their database, or all of it if there is enough space, should be run from memory in the new SQL. Although the processes run from within memory, log files and other tracking data still are periodically written to disk, though this is not part of a normal transaction, and should not slow down those processes any. According to Microsoft, Hekaton should speed up processes by as much as 50 times.
Bottom line: The Hekaton technology would be another tool in the belt of SQL Server database administrators and should appeal to folks in situations where speed or massive data transactions are key. As big data processes grow in the public sector, in-memory computing will become increasingly important. It gives smaller and mid-size database admins the same type of tools found in larger enterprise setups, and thus, probably can’t come soon enough.
John Breeden II directs the GCN Lab.