Fakultät für Informatik TU München - Fakultät für Informatik
Lehrstuhl III: Datenbanksysteme
Technische Universität München
Home  |  Personen  |  Forschung  |  Lehre  |  Sonstiges  | 

Database Patchwork on the Internet

Contents

Overview

The core of our system is a distributed query engine, which is capable of incorporating services of three different service providers: The integration of data providers means that wrappers for arbitray data sources can be loaded dynamically into a running query processor. Through the use of these wrappers, the data sources are also included dynamically.

Functionality can be included in several different forms:

As with wrappers the integration is performed dynamically by loading the class files from a function provider.

A cycle provider has to run a Java VM executing our server program. Since it is possible to distribute the operators of a query evaluation plan among arbitrary (known) server processes, queries can exploit the new server at once after the registration.

The overall correlations of the different kinds of provider are illustrated in the following picture, which we also used in the form of a poster at the Sigmod 99 conference.

A Travel Agency E-Commerce Portal

We developed some example applications of our system to demonstrate the employment of ObjectGlobe as an enabling technology for scalable Internet applications, such as business-to-business (B2B) e-commerce systems. One of the key challenges is to facilitate query processing over the various heterogeneous data sources in order to build integrated product catalogs, match product availability with demand forecasts, or perform price comparisons for procurement.

Our demo includes some examples taken from the field of travel agencies. The different queries, which can be initiated, are further divided in respect to the features of our system they exploit, and the way they access the data sources. Since we use mainly normal WWW data sources, the access to these data sources can be rather slow when the respective Web server is highly loaded or the network connection to this server is congested. Therefore we have materialized parts of the data of theses sources in our local database and for the majority of our example queries there are two versions. One uses the materialized data and the other (with the (online) annotation) uses the Web data.

A sample session shows the interaction with our web-based front-end.


Lehrstuhl für Datenbanksysteme
Letzte Änderung: 25.05.2005 um 14:34:43