Database Models

Is the discussion about NewSQL and NoSQL a re-run of the discussion of SQL and Pre-SQL?

Jim Starkey’s post on the trade-offs of database models (not least that the NoSQL model represents a resurrection of pre-SQL models) has got a lot of interest on the GOOGLE Cloud Computing group and on High Scalability.

The post is well worth reading.  The excerpt below reflects some of the conclusions of someone that has been involved in architecting the internals of databases for over 30 years, with the scars to prove it:

“Over the years I’ve had a professional affiliations with the amorphous data model (Model 204), hierachical (Datacomputer), CODASYL (DBMS-11), and relational (Rdb/ELN, Interbase, Firebird, MySQL, and NimbusDB).  And my friend Tom Atwood started at least half of the OO database companies.  So, if I can’t claim objectivity, I can at least claim in depth personal experience.

Everything is a compromise.  And I deeply believe that the relational model is the best compromise among simplicity, power, performance, and flexibility of implementation.  It does require data description, but so do all other useful database management systems regardless of what it is called.”

NewSQL offers the power of transactional data guarantees and a rich set-based query language (SQL) but in a highly scalable and extremely cloud-friendly form.  And without having to change your corporate skill base, tools, business processes, or application structures.  As Jim says we’re in a brave new world for data management.

Not that NuoDB has an aversion to Key/Value stores in general.  In fact the back-end storage of the NuoDB database (the “Archive Nodes”) are Key/Value stores.

NuoDB is in Beta 1 release.  If you’re interested in checking it out the Beta application is here.

Add comment

Log in to post comments