When it comes to the many forms of Big Data these days, it seems that the conventional wisdom of the day is that something other than SQL is going to be required. But there are those who would argue that the real issue isn't SQL itself, but rather the limitation of existing relational database engines.
Michael Stonebraker, CTO for VoltDB, a provider of a relational database system that runs in memory, says IT organizations need to start distinguishing between old and new SQL. Stonebraker, who led the development of the Ingres and Postgres databases, says that memory is the new disk, so when it comes to handling large amounts of data using SQL, IT organizations need to look for new database engines that support higher-level languages such as SQL.
The issue with NoSQL approaches, argues Stonebraker, is that they require IT organizations to work with alternative low-level interfaces that are difficult to manipulate. As a result, Stonebraker says any approach that gives up on SQL is simply misguided.
The reason that this is becoming a hot-button issue is because IT organizations have invested billions of dollars in investments in SQL. Adding new data management frameworks such as Hadoop will add considerable expense in terms of finding people with the skills needed to manage these platforms. Stonebraker isn't necessarily against Hadoop; he's just pointing out that there is no one SQL database engine that fits all requirements and that before IT organizations adopt a NoSQL approach, they should consider other SQL-compatible approaches to solving the same problem.
As part of that effort, Stonebraker says that when it comes to business intelligence and analytics applications in particular, databases that provide a columnar format, versus a row-level format, are fundamentally faster given the performance limitations of legacy relational database engines. Given the emphasis on these applications, Stonebraker says these applications create another compelling reason to optimize database engines to the application task at hand.
Stonebraker says it's only a matter of time before all these issues combine to unseat the current industry of database leaders. Whether he's right remains to be seen. But one thing that is for certain is that the future of data management, which is the subject of a webcast hosted by Stonebraker, is never going to be the same again.