While the press/blogosphere and NoSQL advocates will have some fun with Oracle’s apparent flip-flop in this context, the underlying realities, imho, are pretty straightforward: if you need “Web scale” key-value/ISAM-centric storage, Oracle Berkeley DB is a strong competitor (as are Oracle Database and Oracle MySQL, for the vast majority of real-world application scenarios, but that entails digressing from NoSQL dogma); if you need Hadoop-style distributed processing, Oracle can play that card as well as any other vendor (and probably better, with its Exa* products); if you need document-centric storage, Oracle has several options for that domain as well. Check out Oracle SVP Andy Mendelsohn’s Database General Session for more details about the pragmatic realities of what NoSQL means to Oracle and its customers.
Oracle's introduction of its Big Data Appliance at the OpenWorld conference here this week is an indication of the attention it is being forced to pay to NoSQL database technology.
Oracle's appliance is based on Apache Hadoop software, a new Oracle NoSQL database and an open source distribution of R, a programming language for statistical analysis.
The product is being positioned by Oracle largely as something that companies can use to acquire and organize unstructured data before passing it on to its Exadata appliance. Oracle also has tools that will allow companies to run analytics applications directly on the big Data Appliance if they choose to.