Active Databases as Information Systems

1. Active Databasesas Information Systems Dina Q Goldin University of Connecticut, USA with Srinath Srinivasa, IIIT Bangalore, and Vijaya Srikanti, University of Connecticut IDEAS 2004

2. What are Databases? • Databases: systems that manage a large collection of (static) data • Database instance: a state of the database • Users can perform queries & updates • query: current DB state -> query answer • update: current DB state -> new DB state • Integrity constraints • Specify relations between data values in a DB state • Restrict the set of allowable DB states • Databases disallow updates that violate the constraints • passive approach; something more active is desired IDEAS 2004

3. Active Databases • Database system augmented with rule handling • Active approach to managing integrity constraints • ECA rules: event, condition, action • Many other uses have been found for active rules • Maintaining materialized views • Managing derived data • Coordinating distributed data management • Providing transaction models • Etc. • Provably correct universal solutions lacking… • Specifying rules • Rules analysis (termination, confluence, observable determinism) Perhaps the problem is that ADBs should not be viewed as DBs? IDEAS 2004

4. Talk Overview • From Databases to Active Databases • Database Systems vs. Information Systems • The Two Views of Active Databases • as database systems • as information systems • Why switch? IDEAS 2004

5. DBs vs. IS’s Information System = Database + Interaction [GST00] IDEAS 2004

6. Example: Integrity Constraints • Static (Data) Integrity • Express conditions for single database states • Can be specified in first-order logic • Dynamic (Interactive) Integrity - Constraints on state sequences - Express conditions for traversal of state space of the system (i.e. its behavior). - Traversal may involve interaction with user(s). DBs support only static integrity constraints. IDEAS 2004

7. Example: Operations • Operations • Logical unit of dynamics • Services in IS, transactions in DB • IS service: a semantic process that usually involves interaction with one or more external agents (users, other services) • Three types: user, system, prompted • User operations - Initiated by the user (CMS example: submitting a paper). - System cannot predict when it is initiated, nor its trajectory. • System operations • Initiated and carried out by the system (ex: deleting expired member). • Internal to the system; user does not control, may be unaware. • Prompted operations • Initiated by the system, involve interaction with the user (CMS example: assign reviewer to paper after it’s submitted). DBs support only atomic user operations. IDEAS 2004

8. The Principles of the Two Fields • We have provided a working distinction for the purpose of discussing the “principles” of each field • Borders between DBs and ISs are blurred: • Database Systems provide a data management service to users (can be viewed as a simple information system) • Data-Intensive Information Systems perform data management • Active DBs fall within that blurry area • a DB augmented with active rule handling (to perform system operations) • a data-intensive IS restricted to rule-handling services IDEAS 2004

9. Two Views of Active Databasesas Databases with Rules as specialized IS The traditional DB view is more limiting, does not allow ADBs to achieve their full potential. IDEAS 2004

10. Talk Overview • From Databases to Active Databases • Database Systems vs. Information Systems • The Two Views of Active Databases • as database systems • as information systems • Why switch? • Active views example • ADB wish list IDEAS 2004

11. ADBs as Information Systems • The IS view naturally endows ADBs with needed functionality • Individualization: customization of requests, as well as of feedback viewing • Use of logs and profiles • User interaction • Interaction among rules • Dynamic integrity • Example: active views IDEAS 2004

12. Active Views • Active Views • Mediate between ADB users and data • Automate user monitoring of certain events or condition • Provide subscription services and notification services • DB view of active views - Compile user specifications into active rules. - Provide active caching of materialized views. • IS view of active views • Provide individualized services to ADB users: event monitoring, subscription and notification. • A more natural approach IDEAS 2004

13. ADB Wish List • Rule instances • Support multiple instances of the same rule • Now possible only when the condition part of their ECA structure differs. • Can be directly mapped to different instances of IS services. • Rule history • Store the history of events, conditions, actions for each rule instance. • To help transactions handle dynamic integrity violations during rule execution. • Rule interaction • Allow rules to enable, disable, or wait for other rules. • As separate functionality rather than by extending the condition part of ECA structure. • Rules need not be aware of external control over their behavior. • For easier formulization of synchronization across semantic services. IDEAS 2004