The Database and Information System Research Group at the University of Ulm [original]

The Database and Information System Research Group

at the University of Ulm

Peter Dadam, Wolfgang Klas

University of Ulm

Faculty of Computer Science

Dept. Databases and Information Systems

D-89069 Ulm, Germany

E-mail: {dadam, klas}@informatik.uni-ulm.de

WWW: http://www.informatik.uni-ulm.de/dbis

1. General Overview

The University of Ulm was founded in 1967 with focus

on medicine and natural sciences. In 1989 the University

established two new faculties: Engineering Sciences and

Computer Science. This enlargement took place within

the framework of the so-called Science City Ulm. In a

joint effort, the State of Baden-Württemberg, industrial

companies, the University, and the City of Ulm success-

fully established a research and development infrastruc-

ture at or nearby the university campus consisting of the

university's research labs, university-related research

institutes like the Research Institute for Applied Knowl-

edge Processing (FAW), and industrial research and de-

velopment labs, especially a large research center of

Daimler-Benz AG.

Today, the Faculty of Computer Science consists of seven

divisions (called 'departments'), each of which equipped

with two professor positions:

- Theoretical Computer Science

- Artificial Intelligence

- Distributed Systems

- Databases and Information Systems

- Software Technology and Compiler Construction

- Computer Structures

- Neural Information Processing.

The Dept. of Databases and Information Systems (DBIS)

became operational at the beginning of 1990 when Peter

Dadam joined the faculty. He came from the IBM Heidel-

berg Science Center (HDSC) where he managed the re-

search department for Advanced Information Management

(AIM). At the HDSC he was working on advanced

database technology and applications and contributed to

the development of the AIM-P system (see [1]). The

second professor position was first occupied by Marc

Scholl, who belonged to the DBIS department from 1992

to 1994. In 1996 Wolfgang Klas joined the DBIS depart-

ment as second professor. He came from the GMD Insti-

tute for Integrated Publication and Information Systems

(IPSI) where he managed the research division Distrib-

uted Multimedia Information Systems and was working on

advanced object-oriented database systems technology,

interoperable database systems, and multimedia informa-

tion systems.

At present, the DBIS team consists of the teaching and

research assistants Thomas Bauer, Susanne Boll, Chris-

tian Heinlein, Clemens Hensinger, Erich Müller, Manfred

Reichert, Birgit Schultheiß, the system engineer Rudi

Seifert, the secretary Christiane Köppl, and the doctoral

students Thomas Beuter and Anita Krämer.

In the following, we concentrate on the research and

development work performed previously and presently in

the research groups of Peter Dadam and of Wolfgang

Klas. For references to Marc Scholl's work please visit

http://www.informatik.uni-konstanz.de/dbis.

2. Database Technology and

Advanced Database Applications

2.1 Database Technology

Most of the research in this area was performed from

1990 to 1995. To a certain degree it was a logical con-

tinuation of the research activities performed in the

AIM-P project at the HDSC. The two major areas were

- Query optimization and support of flexible storage

structures in extended NF2 DBMS

Today's object oriented DBMS as well as most of the

research prototypes based on the NF2, extended NF2, or

similar data models usually provide a (more or less)

hard-wired mapping from logical to physical storage

structures. The research performed in this field concen-

trated on the issues of (a) achieving a higher degree of

data-independence by supporting a variety of different

physical storage structures for a given logical object

structure, (b) finding rules for the utilization of (path)

indexes during query processing, and (c) developing

appropriate query optimization and execution tech-

niques to utilize these features.

- Support of variant structures in relational and NF2

relational database management systems

Today’s relational, object-oriented DBMS, and research

prototypes for complex object support do not provide

appropriate mechanisms to adequately support variant

data structures. Neither null-values nor attribute inheri-

tance in object-oriented data models do really solve this

problem. The challenge was to develop a generalized,

extended relational data model, to redefine the rela-

tional algebra operators accordingly, and to understand

the consequences for query execution (cf. [3] for more

details).

2.2 Advanced Hospital Information Systems

We consider advanced hospital information systems as

being one of the most challenging and interesting applica-

tion areas for databases, workflow technology, and soft-

ware development in general. Very often, especially in the

context of university hospitals, we find rather decen-

tralized organizational structures with (at least partially)

decentralized software development and maintenance

leading to "information islands". We find personnel

working under (partially extremely) high time pressure,

who must often make important decisions about patient

treatment within a short period of time. We find strictly

predefined medical procedures which have to be obeyed

on the one side, and on the other side find many cases and

circumstances in which ad-hoc deviations from these

procedures are mandatory (by nature this probably

happens much more frequently than in any other

application area). – The DBIS department, therefore, has

been engaged in clinical projects almost from the very

beginning.

From 1992 to 1995 the DBIS department was leader of

the large interdisciplinary research project Open Clinical

Database and Information System for the Integration of

Autonomous Subsystems funded by the State of Baden-

Württemberg. Within this project several clinics (internal

medicine, clinical chemistry, anesthesiology), the hospital

computing center, and the DBIS department worked

closely together to identify the technological requirements

with respect to adequate clinical information systems. The

(initial) goal of this project was the development of con-

cepts for the provision of information and application

services in a heterogeneous, open systems environment

(see [4]) with existing information islands, decentralized

software development and thus decentralized provision of

electronic services.

During the project it turned out, however, that although

improved access to patient data is very important, it will

not sufficiently solve the core problems in the clinical

routine. The reason is that today physicians are already

confronted with a massive load of data which have to be

intellectually processed and structured. As this usually has

to be done within a rather short period of time this leads

to many kinds of mistakes. An advanced computer-based

system should therefore not only provide on-line access to

patient data, but should also relieve medical personnel

from tracking the state of medical activities like proc-

essing of a medical order, and from caring about the

timely collection of their results.

In the sequel, we have therefore concentrated on concepts

for process-oriented (clinical) application systems which

actively support the medical personnel in keeping track of

events, in scheduling and in coordinating tasks, and in

cooperating with other clinical staff. The challenge was

on the one side to understand in-depth all characteristic

types of processing, the organizational structures, the

kinds of exceptions which may occur in clinical proc-

esses, and the adequate reaction on such events, to learn

how computer support can be smoothly integrated in the

daily routine work, and how adequate man-machine inter-

faces for clinical environments have to look like. – The

challenge on the other side was to find the adequate tech-

nological basis for supporting these requirements (see

Sect. 3).

On the application side we are working into two direc-

tions: "Smooth assistance" of routine work by knowledge-

based components on the one side, and business process

reengineering and workflow management (the larger

activity) on the other side. With "smooth assistance" we

mean that in clinical application environments the com-

puter may only make suggestions but is never allowed to

make final decisions. This is especially true when deci-

sions about patient treatment have to be made (because of

moral but also of legal reasons). In order to understand

how such an assistance could look like, we have devel-

oped a component to assist the physician in selecting tests

and medical interventions based on medical guidelines

which suggests (further) tests and treatments for a given

indication and on outcomes obtained so far (see [2]). –

Similar rules could be used in principle to describe inter-

workflow dependencies on a logical level. These rules

could be used as a basis for generating the input for an

inter-workflow execution monitor (see Sect. 3.3).

At the beginning of 1996, we started the two-years project

Using Workflow Management Systems for Clinical

Applications in cooperation with the Women's Hospital of

the University of Ulm and the workflow division (Work-

Party development) of Siemens-Nixdorf AG. The goal is

to elaborate a complete "process map" for this hospital,

and to evaluate to which extent today's workflow technol-

ogy – especially the workflow management system

(WfMS) WorkParty – is (already) able to meet the re-

quirements of clinical workflow applications.

Within this project we have thoroughly analyzed the or-

ganizational structures and deeply analyzed all relevant

(business) processes of this hospital. Some of them have

been redesigned and optimized, which has partially led to

significant improvements. To gain concrete implementa-

tion and usability experience we have implemented the

complete business process (including the integration of

foreign systems) for the division "day clinic" (which per-

forms minimal invasive surgery) of this hospital based on

WorkParty. We have completely replaced the standard

user interfaces of the underlying WfMS by own applica-

tion-oriented end-user interfaces, and we have also

extended the capabilities of the WfMS by the provision of

notification services (making the server an "active"

component), the support of time, and "semantic rollback"

across several steps (see also Sect. 3).

We expect that the clinical domain will continue to be one

of our major areas of application-oriented research and a

good test bed for our research work in cooperative infor-

mation systems and workflow technology.

2.3 Concurrent Engineering

Since 1993 we have closely cooperated with the Daimler-

Benz Research Center in Ulm in the field of Concurrent

Engineering. The ultimate goal is to improve the product

development process, i.e., to shorten development times,

to maintain or improve product quality, and to reduce

costs. Within this process we are contributing to the iden-

tification and, where necessary and appropriate, to the

development of the adequate technological basis to per-

form this task.

3. ADEPT♣: Cooperative Information

Systems and Workflow-Management

3.1 Dynamic Changes

One of the most severe weaknesses of today’s workflow

technology is the absence or very limited ability to deviate

(in a controlled and secure way) from the pre-planned

process template at run-time. For many clinical processes

(except some simple ones) it is usually either not possible

or at least not cost-effective to put all possible deviations

from the "standard plan" as possible execution sequences

into the process template. We are therefore working to-

wards workflow technology which allows to deviate from

the pre-modeled process template (skipping of steps,

going back to previous steps, inserting new steps, etc.) in

a secure and safe way. That is, the system guarantees that

all consistency constraints (e.g., no cycles, no missing

input data when a task program will be invoked) which

have been ensured prior to the dynamic ("ad hoc")

modification of the process instance are also ensured after

the modification (see [5], [6]).

3.2 Temporal Aspects

Today’s WfMS usually support the concept of time rather

rudimentarily. Deadlines for the execution of steps can be

specified, but the temporal dependencies between them

are usually not known to the system. Thus, the conse-

♣ ADEPT stands for Application Development Based on

Encapsulated Premodeled Process Templates.

quences for missing a certain deadline with respect to

subsequent steps are not directly inferable. The experi-

ences with clinical applications (see Sect. 2.2) have led to

the conclusion that the support of time and time-depend-

encies are very important features of a WfMS to become

applicable for a broader range of applications. By doing

so, WfMS are going to provide functionality usually

found in calendar and scheduling systems. – This func-

tionality especially becomes important in the context of

supporting dynamic changes as described in the previous

section.

3.3 Inter-Workflow Dependencies

Many real-world tasks – not only in the clinical context –

are interrelated in the one way or another. The tasks "en-

doscope examination of the stomach" and "ultrasound

examination of the stomach", for example, are interrelated

in at least two ways: Normally, they are performed in

different examination rooms and can therefore not be

executed for the same patient at the same time. In addi-

tion, they also interfere with one another. If the endoscope

examination (which requires to fill the stomach with air)

is performed at first, the ultrasound examination cannot be

performed on the same day. At first glance, it would

therefore make sense to combine such interrelated tasks at

the modeling level into one single workflow template

(process template) which would, in principle, open the

possibility to obey all the dependencies by ordering the

steps within the overall workflow accordingly. However,

this approach causes problems for several reasons: At

first, the two tasks need not to be initiated at the same

time. That is, each possible interleaving of steps may

occur at execution time and would have to be taken into

account at the modeling level. In addition, it is not suffi-

cient to consider the tasks in pairs. Every combination of

interfering tasks – with all possible interleaving – would

have to be considered. Obviously, this would lead to a

huge number of workflow templates, each of which very

complex and hard to maintain. Therefore, mechanisms

have to be developed which allow to express inter-

workflow dependencies and to monitor and steer the exe-

cution of such workflow instances at run-time.

3.4 Further Activities

Further activities in this area are falling into the cate-

gories:

- Component-based software development

- Environments for large-scale workflow executions

- Workflow scheme evolution

- Modeling concepts

- Man-Machine interfaces

- Architectural issues and ADEPT implementation

4. Multimedia Information Systems

4.1 Multimedia Repositories

Our work in the area of multimedia information systems

focuses on the integrated management and presentation of

multimedia data by a DBMS. The overall approach taken

is driven by a project aiming at the development of a

network-based and database-driven multimedia informa-

tion system for doctors, medical lecturers, students, and

patients. The application requirements are provided by the

University Hospital of Ulm, its Cardiological Clinic,

Surgical Clinic, and the associated Rehabilitation Hospi-

tal. In this context, we are developing concepts and proto-

typical implementations of a multimedia data repository

that supports

- efficient management of static multimedia material for

courses in the domain of heart surgery,

- dynamic multimedia presentations including photo-

graphs, slides, X-ray images and findings, video record-

ings of heart surgeries, and audio annotated material,

- interactive case-based training and teaching material,

- patient-oriented information.

The conceptual and technical challenges arise from the

approach to seamlessly integrate flexible dynamic multi-

media presentation services with the multimedia reposi-

tory. We are concerned with the database-driven genera-

tion of multimedia presentations depending on output

channels, e.g., university campus-based or home-based

channels, on the actual performance of infrastructure as

well as on the presentation quality and content requested

by the consumer. This calls for a presentation-neutral

representation of multimedia material in the database,

e.g., representing text as SGML documents (e.g., [7]),

providing images, audio, and video data in different

presentation quality, and allowing to store modular

presentation plans independently of the final presentation

format. The system has to support the dynamic

composition of fragments of such presentation plans ac-

cording to user requests. That is, the repository will con-

tain pre-orchestrated "presentation beans" which can be

composed to final presentations on the fly. The quality of

a presentation may still be a parameter for such a compo-

sition of presentation beans and will be determined by the

output channel chosen by the user. For example, a high

quality presentation dynamically composed during a

seminar at the university campus should be available for

students at home although they do not need the high

quality of videos or images at home. This obviously re-

quires a configurable presentation execution environment

as the front end of the repository. The repository has to

provide rich modeling capabilities including the repre-

sentation of metadata for efficient indexing of multimedia

material [9] and it has to support a variety of standards

including, e.g., the medical standard DICOM (Digital

Imaging and Communications in Medicine).

The approach taken in this project aims at the provision of

a generic presentation service by the multimedia reposi-

tory [8, 10, 11, 12, 13]. It is based on the AMOS project

which initially started at GMD-IPSI, and it makes use of

object-relational database technology. The prototype is

being implemented in Java.

4.2 Internet-based Information Systems

In the context of an international project sponsored by the

European Commission we are currently prototyping a

database-driven electronic commerce system, the Elec-

tronic Market Place (EMP). With this system producer

and consumer in rural and remote regions can participate

in the market at the same conditions as participants in the

market in urban regions. The electronic market place

allows to execute business transactions, i.e., to offer and

to buy goods and services without any intermediary and

wholesalers. The market trades not only arbitrary goods

but offers as well services like financing and logistics.

These services are integrated in the traditional buying and

selling services. All product and service data is safely and

consistently managed by a central instance, the EMP

server. The traders connect to the EMP server and place

offers or retrieve offers from there.

The realization of the project is based on existing infra-

structure such as WWW and Internet. But it will still be

possible to participate with traditional communication

channels such as fax, phone, and mail. The major issues

addressed in our group are the design of the database

server which handles all kinds of media objects, the de-

sign and integration of the EMP services with the data-

base server, and the realization of these components in

Java.

5. Other IS-related Research

The group of Michael Weber, Dept. for Distributed

Systems (see http://www-vs.informatik.uni-ulm.de),

pursues research closely related to the DBIS activities.

They focus on system support for cooperative applications

and collaborative scenarios. Specific research projects are:

- Middleware for integrated cooperative systems

The concepts of brokerage and trading in distributed

systems have been extended to seamlessly bind asyn-

chronous (e.g., workflow management) and synchro-

nous (e.g., desktop conferencing) CSCW systems. This

approach combines techniques known from distributed

object computing with methods from agent technolo-

gies. These are augmented by concepts tailored for

multi-user, cooperative applications. Proofs of these

concepts have been demonstrated in real life scenarios

taken from civil authorities. A dedicated trial is cur-

rently being conducted with building constructors and

architects.

- Collaboration support for teleteaching

In a joint project with members of various faculties a

teleteaching framework is being developed. While other

partners participate in the role of content providers, the

group of Michael Weber is enhancing the framework by

customizable and adaptive collaboration support. A

range of components including audio/video conferen-

cing, discourse system, help desk and lightweight work-

flow management is offered. All components can be

linked to the content using Java applet technology.

- WWW-based call center for sensitive medical

applications

This generic call center provides Internet remote access

to patient databases being set up for specific diseases or

syndromes. A help desk extends the plain database

retrieval by means of tele-consultation. Major research

issues are security and privacy in such Internet-based

applications.

- Distributed multimedia information systems

Together with media authors and screen designers a

municipal information system has been developed. This

system uses an ATM broadband network in town and

slow lines to connect to the Internet. Software archi-

tectural issues concerning the diverse network capabili-

ties while ensuring fast response times and artistic

screen design have been the focus of this work.

6. Infrastructure

For our teaching as well as for our research we consider it

very important to teach not just concepts and to work not

only at a conceptual level, but also to have hands-on

experience with existing systems. We, therefore, operate a

large variety of systems in our department. At present we

have installed the following commercial systems:

Hardware: Sun, RS/6000, PCs

Operating Systems: Solaris, AIX, Novell, OS/2, Win

3.11, Windows NT

Database Management Systems: Oracle, DB2, O2,

ObjectStore, Illustra/Informix with various multimedia

and web extensions

Modeling Tools: Aris-Toolset, Aeneis, Bonapart,

LeuSmart, ProMod, Statemate

Workflow Management and Groupware Systems:

Aris-Workflow, FlowMark, Groupflow, Lotus Notes,

ProMInanD, WorkParty

Other Software: e.g., CICS/6000, openUTM, Encina,

DCE

7. References

[1] Dadam, P., Linnemann, V.: Advanced Information

Management (AIM): Advanced Database Technology

for Integrated Applications, IBM Systems Journal, Vol.

28, No. 4, 1989, pp. 661-682.

[2] Heinlein, C.; Kuhn, K.; Dadam, P.: Representation of

Medical Guidelines Using a Classification-Based

System. Proc. CIKM ’94, Gaithersburg, Maryland,

Nov./Dec. 1994, pp. 415-422.

[3] Kalus, C., Dadam, P.: Flexible Relations - Operational

Support of Variant Relational Structures, Proc. VLDB

‘95, Zürich, Sept. 1995, pp. 539-550.

[4] Kuhn, K.; Reichert, M.; Nathe, M.; Beuter, T.; Dadam,

P.: An Infrastructure for Cooperation and Communica-

tion in an Advanced Clinical Information Systems.

Proc. 18th Ann. Symp. on Computer Applic. in Medical

Care 1994, Washington, 1994.

[5] Reichert, M.; Dadam, P. ADEPTflex – Supporting

Dynamic Changes of Workflows Without Loosing

Control. Univ. of Ulm, Faculty of CS, Res. Rep. No.

97-07, April 1997.

[6] Reichert, M.; Dadam, P.: A Framework for Dynamic

Changes in Workflow-Management Systems. Proc. Int'l

Workshop on Database and Expert Systems Applic.,

Toulouse, France, Sept. 1997, pp. 42-48.

[7] Böhm, K., Aberer, K., Klas, W.: Building a Hybrid

Database Application for Structured Documents,

Journal on Multimedia Tools and Applications, 5,

1997, Kluwer Academic Publishers, pp. 275-300.

[8] Boll, S., Klas, W., Löhr, M.: Integrated Database

Services for Multimedia Presentations. In: S.M. Chung

(ed.), Multimedia Information Storage and

Management. Kluwer Academic Publishers, 1996.

[9] Boll, S., Klas, W., Sheth, A.: Overview on Using

Metadata to Manage Multimedia Data. In W. Klas, A.

Sheth (Eds.): Managing Multimedia Data: Using

Metadata to Integrate and Apply Digital Data,

McGraw-Hill, 1997.

[10] Boll, S., Wäsch, J.: A Java Application Programming

Interface to a Multimedia Enhanced Object-Oriented

DBMS. Proc. First Int'l Workshop on Persistence and

Java (PJ1), Glasgow, Scotland, Sept. 1996.

[11] Klas, W., Aberer, K.: Multimedia and its Impact on

Database System Architectures. In: P. Apers, H.

Blanken, M. Houtsma (eds.): Multimedia Databases in

Perspective, Springer Verlag Berlin/Heidelberg, 1997.

[12] Thimm, H., Klas, W., Cowan, C., Walpole, J., Pu, C.:

Optimization of Adaptive Data-Flows for Competing

Multimedia Presentational Database Sessions, IEEE

Int'l Conf. on Multimedia Computing and Systems,

June 1997, Ottawa, Canada.

[13] Vazirgiannis, M., Boll, S.: Events in Interactive

Multimedia Applications: Modeling and Implemen-

tation Design, IEEE Int'l Conf. on Multimedia Com-

puting and Systems, June 1997, Ottawa, Canada.