Lehrgebiet Informationssysteme

Enhanced database applications primarily request for short response times on complex queries from DBMS. Therefore, enhanced DBMS which implement new data models have to exploit parallel query processing strategies.
In the PRIMA-project (prototype implementation of the MAD-data model), we investigate parallel query processing in a complex-object DBMS. We initially con centrate on closely coupled multiprocessor systems, because manifold relationships between elementary objects complicate adequate distribution of objects in a distributed system. In PRIMA, a functional decomposition led to a client/server architecture, which has been implemented on a system-specific runtime system, called Remote-Cooperation-System. Compared to conventional (e.g. relational) DBMS, increased modeling capabilities and different processing characteristics, offer new promising opportunities for parallel query processing. In this project, we concentrate on the following topics:

efficient runtime system support for our client/server implementation.
load distribution of requested tasks,
design and implementation of parallel Algorithms for complex-object construction from elementary objects,
design and evaluation of cost functions for effective query optimization, and finally,
integration of these resulting cost model into PRIMA.

A couple of special measurement tools allow for system evaluation on different layers (overall system, server, algorithm).