The hot ticket to database performance these days is “in-memory” and many are now exploring how this can improve their data analytics capabilities. The basic premise is to load all critical data directly in RAM and conduct lightning fast processing there. As main memory is many times faster than spinning disk, even faster than flash storage, this method has an inherent physics advantage. The cost of doing this has come way down and organizations are seeing that boost translate into real-time, or near-real-time, insights even on large data sets. Various vendors have designed this natively into their products or are retrofitting this performance into their databases. microsofts-surefire-big-data-strategy-run-what-you-brung/index.html" target="_blank">(See my post on Microsoft last week.)
has taken a pure-play approach with their distributed SQL offering, and has shown fantastic results with a simple but well-optimized scale-out model. The 3.0 release added columnar-indexing and an automated hierarchical approach to storage tiers, along with JSON functionality. This all has just gotten better with the addition of a slick operations dashboard for the poor, over-worked DBA who has to manage and tune any size environment. Monitoring, planning, and troubleshooting is facilitated with an intuitive and customizable graphical user interface.
While not alone in the in-memory sub-segment of the SQL database market, this shows increasing maturity from an operations point of view, and that is sorely needed in the big data market in general. Keeping everything in memory shouldn’t be at the expense of management tools, and new innovative software needs exactly this sort of enhanced functionality to have a sporting chance against the bigger database incumbents like
Microsoft’s SQL Server 2014
IBM’s DB2 with Blu Acceleration
*Did anyone actually see Johnny Mnemonic and get the reference? For the record, William Gibson’s writing is way better than the film would indicate.