Keywords:

Keywords: replication, data integrity, availability, middleware
Astract: Replication is used in distributed systems to achieve higher availability and/or performance. Correctness criteria for data-centric distributed systems are replica consistency and data integrity (also called constraint consistency). If consistency needs to be ensured all times, such systems soon become (partially) unavailable if node and link failures occur. However, there exist applications (e.g., in air traffic control) in which consistency can be temporarily relaxed during degraded situations in order to achieve higher availability. Thus, consistency can be balanced against availability. This in turn requires repair actions after reunification of network partitions. That is, reconciliation is necessary to re-establish replica consistency and data integrity when the system becomes healthy again.
The main focus of this thesis is on replication techniques for controlling this trade-off in distributed object systems; a secondary focus is on replication techniques for data-centric service oriented systems. Thus, the contribution of this thesis is three-fold:
First, we introduce an enhanced replication model for trading data integrity against availability - the Availability/Consistency Balancing Replication Model (ACBRM).
Second, we present Adaptive Voting - a concrete protocol that realizes the abstract model. Both an analytical availability analysis and a prototype implementation show the feasibility of the approach, especially if (i) a significant portion of data integrity constraints of the system is relaxable and (ii) reconciliation time is shorter than degradation time.
Third, distributed object replication middleware systems (e.g., the DeDiSys middleware which is targeted to the ACBRM) and service replication middleware systems are compared on an architectural level. From this analysis and experiences with our middleware prototypes we conclude that many well-established replication middleware architectures can be reapplied in service oriented systems.
Future work needs to focus on dependability and security techniques for service oriented systems of the future characterized by cross-organizational heterogeneity, massive scale, and dynamicity.

Keywords: Ubiquitous Web Applications, Aspect-oriented Modeling, Model-Driven Software Development, Web Modeling
Astract: Ubiquitäre Web-Anwendungen (UWA) stellen speziell auf die aktuelle Situation angepasste Informationen und Dienste zur Verfügung. Diese Anpassung an den aktuellen Kontext, z.B. an verschiedene Endgeräte, Benutzer mit diversen Interessen, deren Aufenthaltsort sowie zeitliche Aspekte, wird als Customization Funktionalität bezeichnet. Die Entwicklung einer UWA unter Berücksichtigung von Customization, die sich quer durch die Kontent-, Hypertext- und Präsentations-Ebenen einer Web-Anwendung zieht, gestaltet sich jedoch äußerst komplex und bedarf einer systematischen Entwicklung auf Basis von Modellen im Sinne des Model-driven Engineerings (MDE). Dabei verspricht MDE effizientere Entwicklungsprozesse sowie bessere Wartbarkeit und Weiterentwicklung von Software durch die semi-automatische Generierung von Software aus Modellen. Im Modell einer Web-Anwendung ist Customization jedoch inhärent mit der Kernfunktionalität vermischt und behindert dadurch die Verständlichkeit, Wiederverwendbarkeit, Wartbarkeit und Weiterentwicklung. Die Aspekt-Orientierung bietet dafür einerseits neue Konzepte um Querschnittsfunktionalität wie Customization in so genannten Aspekten zu modularisieren und andererseits die notwendigen Mechanismen für die Integration dieser separierten Funktionalitäten, um ein verwendbares Gesamtmodell zu erhalten. Einige wenige Web Modellierungssprachen unterstützen bereits die aspekt-orientierte Modellierung (AOM) von Customization jedoch mit folgenden Restriktionen:
Erstens wird Customization nur für die Hypertext-Ebene, getrennt von der Kernfunktionalität der Web-Anwendung behandelt. Zweitens werden nur wenige grundlegende Konzepte der AOM eingesetzt, was in limitierten Mechanismen zur Trennung von Customization resultiert. Drittens, wird die spätere Integration der Aspekte in Modelle nicht unterstützt. Und viertens ist die model-getriebene Entwicklung einer UWA im Sinne von MDE aufgrund fehlender Metamodell-Spezifikationen und Werkzeugunterstützung oft nicht möglich.
Das Ziel der vorliegenden Dissertation ist der umfassende Einsatz von AOM Konzepten, um einerseits den Customization Aspekt auf allen Ebenen einer Web-Anwendung getrennt modellieren und andererseits den Aspekt auch wieder mit der Kernfunktionalität integrieren zu können. Als Lösung wird der aspectUWA Ansatz vorgestellt, welcher die Erweiterung existierender Web Modellierungssprachen um AOM Konzepte auf Basis eines generischen Rahmenwerks diskutiert. Die Anwendung des aspectUWA Ansatz wir anhand der Web Modellierungssprache WebML gezeigt. Dabei sind die wesentlichen Beiträge dieser Dissertation: (i) Das so genannte Conceptual Reference Model als generisches Rahmenwerk für die Erweiterung von Web Modellierungssprachen um aspekt-orientierte Modellierungskonzepte. (ii) Eine Metamodell-Spezifikation für WebML, welche semi-automatisch aus einer existierenden Sprachspezifikation auf Basis von DTDs entwickelt wurde. (iii) Die aspectWebML Sprache, die auf dem konzeptuellen Referenzmodell und dem WebML Metamodell aufbauend entwickelt wurde und die Modellierung von Aspekten als auch deren Integration unterstützt. (iv) Richtlinien für die Modellierung von Customization mit aspectWebML. Und (v) die aspectWebML Entwicklungsumgebung als Werkzeugunterstützung für die Modellierung und Integration von Aspekten mit aspectWebML.

Keywords: UML 2.0, Java, Enterprise applications, Rational Unified Process
Astract: Today not only the world but also the requirements for computer-based applications change permanently. In addition to these requirements, the development of technologies and tools also continues.
Modern and object-oriented technologies, such as UML and Java meet today's high standards. Furthermore, the object-oriented programming language Java, or more precisely the Java Enterprise Edition, is the perfect tool to develop distributed systems including web applications. Also the object-oriented modelling language UML is able to keep up with the rapid development. This thesis evaluates the interaction between UML 2.0 and Java Enterprise applications. The language UML 2.0 offers a total of 13 different classes of diagrams.
Each of these diagrams presents some strengths but also some weaknesses in relation to Enterprise applications. This thesis shows which parts of an Enterprise application can be modelled using which kind of UML 2.0 diagram as well as their incompatibilities. Besides the interaction between UML 2.0 and a web application developed by means of Java Enterprise technology, this thesis also presents the possibilities to model general aspects of an Enterprise application, such as design patterns and system architectures. In this process, all available diagrams offered by UML 2.0 for a given example are modelled. In addition, an overview of the various diagrams in UML 2.0 in the Rational Unified Process is a further objective of this work. In this realm, the question of which kind of diagram is applicable in which phase of the unified process is evaluated. The unified process will be examined from an Enterprise application point of view.

Keywords: algorithms, self-organization, networks, distributed systems, peer-to-peer, folksonomies, social bookmarking, user profiles, visualization
Astract: Peer-to-peer networks and folksonomies are like living organisms, ever growing and changing as time goes on. This thesis addresses the applicability of algorithms derived from the self-organizing and emergent behavior observed from ant colonies to these complex networks for two specific purposes, namely (1) content-based search in unstructured peer-to-peer networks, and (2) the extraction of adaptive user profiles from folksonomies.
For search in unstructured peer-to-peer networks, the main goal is to find the shortest path from every querying peer to one or more answering peers that possess resources which are appropriate answers for the given query. The SemAnt algorithm, which is designed for this task, is based on reputation learning. In reputation learning, the information about the remote peers' resources is gained passively by observing the user queries and their answers that are forwarded through the local peer. Every successful query evokes small updates in the routing tables of those peers that are included in the path between the querying and the answering peer. The routing tables are used for subsequent queries to decide which link to follow in order to find appropriate resources.
The SemAnt algorithm is compliant with the Ant Colony Optimization meta-heuristic, and it employs a probabilistic procedure to select outgoing links for query forwarding. This procedure combines an exploiting strategy, which selects those links currently known as the most appropriate ones, with an exploring strategy, which also follows links not currently known as the best ones with the aim of finding better paths not yet explored. A weight defines the ratio between the strategies.
Since the SemAnt algorithm is a content-based approach to peer-to-peer search, its performance depends on how the content is distributed in the network. The evaluation of the algorithm includes an investigation to which extent this is the case. Based on these results, we develop strategies for improvement.
Under the assumption that the resources in the network are annotated according to a taxonomy, and that the query vocabulary is restricted to the leaf topics from the same taxonomy, it is possible to consider also the upper-level topics in the query routing procedure of the algorithm in order to increase its performance.
Ant algorithms include an evaporation feature for integrating a time factor when incrementally creating solutions. This feature is beneficial for the task of learning adaptive user profiles from tagging data. For this purpose we design the Add-A-Tag algorithm, which is based on a combination of an evaporation feature for adapting the user profile to trends over time, and the co-occurrence technique for determining the relationships between tags.
The user profiles created with the Add-A-Tag algorithm are semantic networks derived from the structure of the tagging data, and they are adaptive in the sense that they change according to changes in a user's tagging behavior. In addition to the long-term interests of a user, also his or her short-term interests are included in the profile at any given point of time. We present a tool for visualizing the changes in the profile over time, and we show how to exploit the profile for personalized browsing of annotated data sources.

Keywords:
Astract: This paper presents SemAnt, a distributed content-based routing algorithm based on the Ant Colony Optimization meta-heuristic and adapted for deployment in peer-to-peer networks. Under the assumption that the shared content in the network is annotated according to a taxonomy, it is possible to determine the hierarchical relationships between queries, and to exploit this information to improve the routing process. The experimental results presented in this paper show that the performance of content-based peer-to-peer search is highly dependent on the content distribution in the network and on the network's topology. It can be improved by exploiting the information provided by the underlying taxonomy. The degree of improvement is proportional to the degree of coherence in the content distribution.

Keywords:
Astract: The United Nation's Center for Trade Facilitation and Electronic Business (UN/CEFACT) is a standardization body known for its work on UN/EDIFACT and ebXML. One of its most recent developments is UN/CEFACT's Modeling Methodology (UMM). The UMM standard is used to model inter-organizational business processes in the B2B domain. With the increasing availability of electronic governmental services over networks, the frontier between B2B and B2G/G2G disappears. Today one expects a governmental institution to react like any other business partner. Therefore also governments now face the interoperability and compatibility issues as regular businesses do. In order to allow two governmental institutions to collaborate, a methodology uniquely depicting the inter-organizational process from a global perspective is needed. In this paper we propose to use UN/CEFACT's Modeling Methodology in the eGovernment domain. UMM allows the definition of a global choreography which is then being used to derive local orchestrations for each business partner. Such orchestrations can then be used by enterprise applications in a service oriented context. As an example, a real-world scenario from the waste transport domain within the European Union will be shown. Furthermore the possible integration in the context of the We-Go project [3] is examined.

Keywords:
Astract: Model transformations are playing a vital role in the field of model engineering. However, for non-trivial transformation issues most approaches require imperative definitions, which are cumbersome and error-prone to create. Therefore, Model Transformation By Example (MTBE) approaches have been proposed as user-friendly alternative that simplifies the definition of model transformations. Up to now, MTBE approaches have been applied to structural models, only. In this work we apply MTBE to the domain of business process modeling languages, i.e., Event-driven Process Chains and UML activity diagrams. Compared to structural languages, business process modeling languages cover static semantic constraints, which are not specified in the metamodel. As a consequence, reasoning on the abstract syntax level is not sufficient. The contribution of this paper is to extend our existing MTBE approach by new alignment operators on the user level, which further improves the transparency of model transformation code. Concrete syntax and the knowledge about mapping operators are to be the only requisite artifacts.

Keywords:

Keywords: