Jürgen Zimmermann

Rules in an Open System: The REACH Rule System

REACH is an active object system intended to control heterogeneous systems, possibly under timing constraints. When dealing with open systems in which the controlled system may execute irreversible actions, many notions of active databases must be revised and adapted to this situation. In this paper we draw the system boundaries between controlling and controlled system, present a rule system that includes events and actions both in the controlled and the controlling systems and analyze the effects of this open environment on the rule structure. We identify two new coupling modes, sequential causally dependent and exclusive causally dependent, which are necessary for handling irreversible actions in external systems and contingency plans, respectively.

The REACH active OODBMS

REACH [1] is a full-fledged aOODBMS research platform, that was implemented by extending TI’s OpenOODB [2]. The REACH prototype is one of the first running systems that combines most of the advanced features of aOODBMSs that previously have been shown in partial implementations only. It provides a much needed platform for the development of applications and was instrumented to make performance measurements and visualize the flow of execution. It has a graphical user interface and support tool for rule definition and administration (GRANT). OpenOODB is implemented in an extensible way that resembles the event-driven paradigm of active database systems. A meta-architecture module implements the notions of low-level event, sentry, and policy manager interface. Policy managers are used for handling persistence, transactions, queries, etc. We extended Open OODB with policy managers for event handling and rule execution. REACH implements primitive method events, temporal events, flow control events and the composition operators for sequence, disjunction, conjunction, closure, negation, and history. It uses the C++ type system and detects method events through inline wrapping of the methods in the extended preprocessor of OpenOODB. An ECA rule is mapped to an ECA manager which is accessed during event detection and propagates the event object to the relevant rules implemented as two C functions (for condition and action), which are stored in a shared library [1]. REACH supports a rich set of coupling modes. Rules can be executed either in immediate, deferred or detached mode. Detached mode may be independent or dependent with variants parallel, sequential or exclusive which is used for contingency plans. For details see [1]. Event composition semantics are defined relative to transaction boundaries. Composite events that trigger deferred rules must be composed from events originating in one transaction. Composite events where the constituent events originate in multiple transactions

Design, Implementation and Management of Rules in an Active Database System

Active database systems must provide efficient support for event detection and rule execution. Large applications that benefit from facilities offered by active database systems require tools for their whole life cycle — otherwise there will be no acceptance of this new database technology. Therefore, we propose to extend the well-known and broadly accepted OMT design methodology to model classes and rules which have to be mapped onto a rule language. Then we present the rule languages of NAOS, SAMOS, and REACH and compare them from a users point of view. To improve the rule management REACH offers the facility to organize rules in so-called rule directories which provide a structured name space like UNIX directories for the storage of files. Then we describe a comprehensive set of commands for rules and rule directories, especially an access control mechanism for rules. Additionally browsers for the rules, the rule triggering graph, and the event history are presented, and finally we show how the triggering of rules within applications can be visualized online in a trace mode.

Objective: a benching for object-oriented active database systems

Although much work in the area of Active Database Management Systems (ADBMSs) has been done, it is not yet clear how the performance of an active DBMS can be evaluated systematically. In this paper, we describe the OBJECTIVE Benchmark for objectoriented ADBMSs, and present experimental results from its implementation in an active database system prototype. OBJECTIVE can be used to identify performance bottlenecks and active functionalities of an ADBMS, and to compare the performance of