TU München - Fakultät für Informatik Lehrstuhl III: Datenbanksysteme

The Chair of Informatics III: Database Systems of the Technische Universität München consists of two research groups:

• Prof. Kemper's group focuses on self-organizing distributed information processing.

• Prof. Grust's group performs resarch into database-supported compilers and runtime systems for query (and programming) languages.

Focus of Prof. Kemper's Research Group

Click on image for large version

Prof. Kemper's research group focuses on adaptive information processing technologies in distributed environments. Distribution of the information processing tasks is based on

• Peer-to-Peer (P2P) overlay networks,

• Enterprise Grid Architectures consisting of compute clusters or blade servers, and

• Grid Computing in wide area networks.

The information processing paradigms cover a wide spectrum from

• distributed and extensible database query processing techniques to

• web- and grid-service workflows to

• distributed data stream management.

The administration of widely distributed information processing tasks has to be automated to ensure high performance and fault tolerance. In various projects we develop self-organization techniques that are based on

• Quality of Service-monitoring,

• multi-query optimization to reuse result streams of continuous queries,

• dynamic reoptimization and adaptaion of processing plans, and

• adaptive feedback control utilizing a fuzzy controller.

In the security engineering project we develop a comprehensive authorization policy model for such distributed information processing tasks that cross organizational boundaries.

The figure summarizes the different information processing paradigms and optimization techniques that are developed in the various projects (shown in blue). More details can be found on the corresponding project pages.

Research Focus of Prof. Grust's Group

The pre/post plane, a core concept used by MonetDB/XQuery to implement both, XML encoding and XPath location step evaluation.

Our current research aims to use well-understood relational database system technology

1. to construct efficient and scalable processors for query languages which, originally, are non-relational, and
2. to build query and programming language runtime systems that can cope with very large data volumes (beyond primary memory size).

The current prime outcome of this research is a compiler for the XML query language XQuery. The compiler, dubbed Pathfinder, comes with a code generator that emits relational queries (algebraic operators in DAG-shaped plans of significant size, typically > 100 operators). Our main target database systems is MonetDB, developed by our Dutch partners at CWI, Amsterdam. The pair Pathfinder and MonetDB — known as MonetDB/XQuery — currently defines the gloabl state-of-the-art in scalable database-supported XML processing.

The system is already successfully used in applications such digital forensics, and distributed embedded systems. MonetDB/XQuery development follows an open-source model and you are welcome to try it out and to inject your own ideas.

The compilation techniques built into Pathfinder yield algebraic code for (programming) language constructs such as iteration, conditionals, etc. and should be applicable to languages other than XQuery. Investigations into this direction are underway.

The development of and the research behind Pathfinder is supported by the German Research Council (DFG) under contract GR 2036/2-1.