Scale, Flexibility Place New Demands on Databases
In a new research report, Gartner advises clients to consider the “avant-garde” of new relational databases from vendors like MemSQL, NuoDB, and VoltDB when projects call for large amounts of scalability and elasticity on industry-standard hardware, while retaining the precepts of relational tables and SQL.
Over the past 10 years, the traditional relational database management system (RDBMS) cart hasn’t just been mildly upset—it’s been rammed, flipped over, and its contents distributed to the four winds. The near-simultaneous emergence of mobile devices, cloud platforms, and messy social data, along with the ongoing exponential increase in data volumes, have conspired to expose the limitations of the traditional RDBMS deployed atop a symmetric multi-processor (SMP) scale-up server.
NoSQL database vendors and distributed storage and compute systems like Hadoop have been the headliners in IT’s new world order. Instead of storing modest amounts of relatively structured data in neat little relational tables and accessing it through good old SQL, as traditional RDBMS from IBM, Microsoft, and Oracle have done for decades, we’re now storing huge gobs of semi-structured data in whatever form it arrives in, and adding structure to it only when we access it via special APIs.
NoSQL databases from vendors like MongoDB, Couchbase, and Datastax (Cassandra) are more scalable, more elastic, and more flexible than the RDBMs of old. But those advantages come with compromises, including not adhering to the tight ACID precepts that RDBMS users have come to expect, and introducing new concepts around data persistency and availability. It’s a case of pick your poison.
Now we’re seeing the rise of a crop of new RDBMS vendors that seek to marry some of the capabilities of the NoSQL vendors—namely the distributed nature of the databases that boost scalability and elasticity in cloud deployments—along with the relational table storage mechanism and the full SQL access method that have defined traditional RDBMSs.
This is excerpted from a story that appeared at our sister publication, Datanami. The complete article can be found here.