What’s the Shape of Databases to Come? Mike Carey Myopic Industrial Attendee Bronco Acquisition Corporation (a.k.a. BOracle) For Years We Have Enjoyed The Great Convergence In the beginning there was data And then along came Ted Codd Files – bytes, ISAM, VSAM Prehistoric Databases – hierarchical (IMS), network (CODASYL) Relations and declarative queries Views, query optimization, transactions, ... SQL and the relational database market Resulted in a widely accepted field of knowledge Unquestionably a revolutionary change! Several “Post-Relational” DB Movements Have Failed Object-oriented database systems Object-relational database systems Mostly throughout the 1990’s OO concepts + RDBMS features = OR DBMS Semi-structured (XML) database systems Mid-1980’s to mid-1990’s OO languages + DBMS features = OO DBMS Late-1990’s to early-2000’s Schema-last/schema-never + complex data = XML DBMS None of these movements seriously took hold The World Is Flat, Right? Quoting from the Flat Earth Society homepage: “Enter the Flat Earth Society. For over five hundred years humanity has believed the "round Earth" teachings of Efimovich and his followers. But all hope is not lost. For through all that time, a small but diligent band of individuals have preserved the knowledge of our planet's true shape. And now, after centuries in the Dark Ages, we believe that mankind as a whole is once again ready to embrace the truth that has forever been the Flat Earth Society. Using whatever means are deemed necessary and relying heavily on a callous disregard for the lives and wellbeing of our members, we have slowly but steadily been spreading the news.” “But why? Why do we say the Earth is flat, when the vast majority says otherwise? Because we know the truth.” Let’s Have A Look At Where We Are Today Core business data management Relational DBs & SQL work as designed Applications and application data SAP: tables as transactional storage C++/Java O-R tools: tables as backend stores for mapped language objects MSFT: EDM & EntitySQL for ADO.NET++ MDM: tables as backend stores for product data (for example) with a rich & extensible data model SaaS: tables as backend stores for multi-tenant data sets with tenant-extensible schemas Let’s Have A Look At Where We Are Today (cont.) Relational DBs & SQL not looking so good Glorified record-based file system (VSAM++) Data semantics not captured in the RDB schema Encoded in the application, or Encoded as more data, understood by the application We are losing the data’s model & its meaning! Maybe we need a different modeling approach? Maybe we need domain-specific DBMSs? Or maybe we need <your ideas go here>?