### Refine

#### Year of publication

#### Language

- English (15) (remove)

#### Keywords

#### Institute

- Informatik (15) (remove)

XML (Extensible Markup Language) ist ein sequentielles Format zur Speicherung und Übermittlung strukturierter Daten. Obwohl es ursprünglich für die Dokumentenverarbeitung entwickelt wurde, findet XML heute Verwendung in nahezu allen Bereichen der Datenverarbeitung, insbesondere aber im Internet. Jede XML-Dokumentenverarbeitungs-Software basiert auf einem XML-Parser. Der Parser liest ein Dokument in XML-Syntax ein und stellt es als Dokumentbaum der eigentlichen Anwendung zur Verfügung. Dokumentenverarbeitung ist dann im wesentlichen die Manipulation von Bäumen. Moderne funktionale Programmiersprachen wie SML und Haskell unterstützen Bäume als Basis-Datentypen und sind daher besonders gut für die Implementierung von Dokumentenverarbeitungs-Systemen geeignet. Um so erstaunlicher ist es, dass dieser Bereich zum größten Teil von Java-Software dominiert wird. Dies ist nicht zuletzt darauf zurückzuführen, dass noch keine vollständige Implementierung der XML-Syntax als Parser in einer funktionalen Programmiersprache vorliegt. Eine der wichtigsten Aufgaben in der Dokumentenverarbeitung ist Querying, d.h. die Lokalisierung von Teildokumenten, die eine angegebene Strukturbedingung erfüllen und in einem bestimmten Kontext stehen. Die baumartige Auffassung von Dokumenten in XML erlaubt die Realisierung des Querying mithilfe von Techniken aus der Theorie der Baumsprachen und Baumautomaten. Allerdings müssen diese Techniken an die speziellen Anforderungen von XML angepasst werden. Eine dieser Anforderungen ist, dass auch extrem große Dokumente verarbeitet werden müssen. Deshalb sollte der Querying-Algorithmus in einem einzigen Durchlauf durch das Dokument ausführbar sein, ohne den Dokumentbaum explizit im Speicher aufbauen zu müssen. Diese Arbeit besteht aus zwei Teilen. Der erste Teil beschreibt den XML- Parser fxp, der vollständig in SML programmiert wurde. Insbesondere werden die Erfahrungen mit SML diskutiert, die während der Implementierung von fxp gewonnen wurden. Es folgt eine Analyse des Laufzeit-Verhaltens von fxp und ein Vergleich mit anderen XML-Parsern, die in imperativen oder objekt- orientierten Programmiersprachen entwickelt wurden. Im zweiten Teil beschreiben wir einen Algorithmus zum Querying von XML- Dokumenten, der auf der Theorie der Waldautomaten fundiert ist. Er findet alle Treffer einer Anfrage in höchstens zwei Durchläufen durch das Dokument. Für eine wichtige Teilklasse von Anfragen kann das Querying sogar in einem einzelnen Durchlauf realisiert werden. Außerdem wird die Implementierung des Algorithmus in SML mit Hilfe von fxp dargestellt.

Spatial Queues
(2000)

In the present thesis, a theoretical framework for the analysis of spatial queues is developed. Spatial queues are a generalization of the classical concept of queues as they provide the possibility of assigning properties to the users. These properties may influence the queueing process, but may also be of interest for themselves. As a field of application, mobile communication networks are modeled by spatial queues in order to demonstrate the advantage of including user properties into the queueing model. In this application, the property of main interest is the user's position in the network. After a short introduction, the second chapter contains an examination of the class of Markov-additive jump processes, including expressions for the transition probabilities and the expectation as well as laws of large numbers. Chapter 3 contains the definition and analysis of the central concept of spatial Markovian arrival processes (shortly: SMAPs) as a special case of Markov-additive jump processes, but also as a natural generalization from the well-known concept of BMAPs. In chapters 4 and 5, SMAPs serve as arrival streams for the analyzed periodic SMAP/M/c/c and SMAP/G/infinity queues, respectively. These types of queues find application as models or planning tools for mobile communication networks. The analysis of these queues involves new methods such that even for the special cases of BMAP inputs (i.e. non-spatial queues) new results are obtained. In chapter 6, a procedure for statistical parameter estimation is proposed along with its numerical results. The thesis is concluded by an appendix which collects necessary results from the theories of Markov jump processes and stochastic point fields. For special classes of Markov jump processes, new results have been obtained, too.

Due to the breath-taking growth of the World Wide Web (WWW), the need for fast and efficient web applications becomes more and more urgent. In this doctoral thesis, the emphasis will be on two concrete tasks for improving Internet applications. On the one hand, a major problem of many of today's Internet applications may be described as the performance of the Client/Server-communication: servers often take a long time to respond to a client's request. There are several strategies to overcome this problem of high user-perceived latencies; one of them is to predict future user-requests. This way, time-consuming calculations on the server's side can be performed even before the corresponding request is being made. Furthermore, in certain situations, also the pre-fetching or the pre-sending of data might be appropriate. Those ideas will be discussed in detail in the second part of this work. On the other hand, a focus will be placed on the problem of proposing hyperlinks to improve the quality of rapid written texts, at first glance, an entirely different problem to predicting client requests. Ultra-modern online authoring systems that provide possibilities to check link-consistencies and administrate link management should also propose links in order to improve the usefulness of the produced HTML-documents. In the third part of this elaboration, we will describe a possibility to build a hyperlink-proposal module based on statistical information retrieval from hypertexts. These two problem categories do not seem to have much in common. It is one aim of this work to show that there are certain, similar solution strategies to look after both problems. A closer comparison and an abstraction of both methodologies will lead to interesting synergetic effects. For example, advanced strategies to foresee future user-requests by modeling time and document aging can be used to improve the quality of hyperlink-proposals too.

Mobile computing poses different requirements on middleware than more traditional desktop systems interconnected by fixed networks. Not only the characteristics of mobile network technologies as for example lower bandwidth and unreliability demand for customized support. Moreover, the devices employed in mobile settings usually are less powerful than their desktop counterparts. Slow processors, a fairly limited amount of memory, and smaller displays are typical properties of mobile equipment, again requiring special treatment. Furthermore, user mobility results in additional requirements on appropriate middleware support. As opposed to the quite static environments dominating the world of desktop computing, dynamic aspects gain more importance. Suitable strategies and techniques for exploring the environment e.g. in order to discover services available locally are only one example. Managing resources in a fault-tolerant manner, reducing the impact ill-behaved clients have on system stability define yet another exemplary prerequisite. Most state of the art middleware has been designed for use in the realm of static, resource rich environments and hence is not immediately applicable in mobile settings as set forth above. The work described throughout this thesis aims at investigating the suitability of different middleware technologies with regard to application design, development, and deployment in the context of mobile networks. Mostly based upon prototypes, shortcomings of those technologies are identified and possible solutions are proposed and evaluated where appropriate. Besides tailoring middleware to specific communication and device characteristics, the cellular structure of current mobile networks may and shall be exploited in favor of more scalable and robust systems. Hence, an additional topic considered within this thesis is to point out and investigate suitable approaches permitting to benefit from such cellular infrastructures. In particular, a system architecture for the development of applications in the context of mobile networks will be proposed. An evaluation of this architecture employing mobile agents as flexible, network-side representatives for mobile terminals is performed, again based upon a prototype application. In summary, this thesis aims at providing several complementary approaches regarding middleware support tailored for mobile, cellular networks, a field considered to be of rising importance in a world where mobile communication and particularly data services emerge rapidly, augmenting the globally interconnecting, wired Internet.

Hardware bugs can be extremely expensive, financially. Because microprocessors and integrated circuits have become omnipresent in our daily live and also because of their continously growing complexity, research is driven towards methods and tools that are supposed to provide higher reliability of hardware designs and their implementations. Over the last decade Ordered Binary Decision Diagrams (OBDDs) have been well proven to serve as a data structure for the representation of combinatorial or sequential circuits. Their conciseness and their efficient algorithmic properties are responsible for their huge success in formal verification. But, due to Shannon's counting argument, OBDDs can not always guarantee the concise representation of a given design. In this thesis, Parity Ordered Binary Decision Diagrams are presented, which are a true extension of OBDDs. In addition to the regular branching nodes of an OBDD, functional nodes representing a parity operation are integrated into the data structure, thus resulting in Parity-OBDDs. Parity-OBDDs are more powerful than OBDDs are, but, they are no longer a canonical representation. Besides theoretical aspects of Parity-OBDDs, algorithms for their efficient manipulation are the main focus of this thesis. Furthermore, an analysis on the factors that influence the Parity-OBDD representation size gives way for the development of heuristic algorithms for their minimization. The results of these analyses as well as the efficiency of the data structure are also supported by experiments. Finally, the algorithmic concept of Parity-OBDDs is extended to Mod-p-Decision Diagrams (Mod-p-DDs) for the representation of functions that are defined over an arbitrary finite domain.

Today, usage of complex circuit designs in computers, in multimedia applications and communication devices is widespread and still increasing. At the same time, due to Moore's Law we do not expect to see an end in the growth of the complexity of digital circuits. The decreasing ability of common validation techniques -- like simulation -- to assure correctness of a circuit design enlarges the need for formal verification techniques. Formal verification delivers a mathematical proof that a given implementation of a design fulfills its specification. One of the basic and during the last years widely used data structure in formal verification are the so called Ordered Binary Decision Diagrams (OBDDs) introduced by R. Bryant in 1986. The topic of this thesis is integration of structural high-level information in the OBDD-based formal verification of sequential systems. This work consist of three major parts, covering different layers of formal verification applications: At the application layer, an assertion checking methodology, integrated in the verification flow of the high-level design and verification tool Protocol Compiler is presented. At the algorithmic layer, new approaches for partitioning of transition relations of complex finite state machines, that significantly improve the performance of OBDD-based sequential verification are introduced. Finally, at the data structure level, dynamic variable reordering techniques that drastically reduce the time required for reordering without a trade-off in OBDD-size are described. Overall, this work demonstrates how a tighter integration of applications by using structural information can significantly improve the efficiency of formal verification applications in an industrial setting.

Many real-life phenomena, such as computer systems, communication networks, manufacturing systems, supermarket checkout lines as well as structural military systems can be represented by means of queueing models. Looking at queueing models, a controller may considerably improve the system's performance by reducing queue lengths, or increasing the throughput, or diminishing the overhead, whereas in the absence of a controller the system behavior may get quite erratic, exhibiting periods of high load and long queues followed by periods, during which the servers remain idle. The theoretical foundations of controlled queueing systems are led in the theory of Markov, semi-Markov and semi-regenerative decision processes. In this thesis, the essential work consists in designing controlled queueing models and investigation of their optimal control properties for the application in the area of the modern telecommunication systems, which should satisfy the growing demands for quality of service (QoS). For two types of optimization criterion (the model without penalties and with set-up costs), a class of controlled queueing systems is defined. The general case of the queue that forms this class is characterized by a Markov Additive Arrival Process and heterogeneous Phase-Type service time distributions. We show that for these queueing systems the structural properties of optimal control policies, e.g. monotonicity properties and threshold structure, are preserved. Moreover, we show that these systems possess specific properties, e.g. the dependence of optimal policies on the arrival and service statistics. In order to practically use controlled stochastic models, it is necessary to obtain a quick and an effective method to find optimal policies. We present the iteration algorithm which can be successfully used to find an optimal solution in case of a large state space.

The visualization of relational data is at the heart of information visualization. The prevalence of visual representations for this kind of data is based on many real world examples spread over many application domains: protein-protein interaction networks in the field of bioinformatics, hyperlinked documents in the World Wide Web, call graphs in software systems, or co-author networks are just four instances of a rich source of relational datasets. The most common visual metaphor for this kind of data is definitely the node-link approach, which typically suffers from visual clutter caused by many edge crossings. Many sophisticated algorithms have been developed to layout a graph efficiently and with respect to a list of aesthetic graph drawing criteria. Relations between objects normally change over time. Visualizing the dynamics means an additional challenge for graph visualization researchers. Applying the same layout algorithms for static graphs to intermediate states of dynamic graphs may also be a strategy to compute layouts for an animated graph sequence that shows the dynamics. The major drawback of this approach is the high cognitive effort for a viewer of the animation to preserve his mental map. To tackle this problem, a sophisticated layout algorithm has to inspect the whole graph sequence and compute a layout with as little changes as possible between subsequent graphs. The main contribution and ultimate goal of this thesis is the visualization of dynamic compound weighted multi directed graphs as a static image that targets at visual clutter reduction and at mental map preservation. To achieve this goal, we use a radial space-filling visual metaphor to represent the dynamics in relational data. As a side effect the obtained pictures are very aesthetically appealing. In this thesis we firstly describe static graph visualizations for rule sets obtained by extracting knowledge from software archives under version control. In a different work we apply animated node-link diagrams to code-developer relationships to show the dynamics in software systems. An underestimated visualization paradigm is the radial representation of data. Though this kind of data has a long history back to centuries-old statistical graphics, only little efforts have been done to fully explore the benefits of this paradigm. We evaluated a Cartesian and a radial counterpart of a visualization technique for visually encoding transaction sequences and dynamic compound digraphs with both an eyetracking and an online study. We found some interesting phenomena apart from the fact that also laymen in graph theory can understand the novel approach in a short time and apply it to datasets. The thesis is concluded by an aesthetic dimensions framework for dynamic graph drawing, future work, and currently open issues.

This work addresses the algorithmic tractability of hard combinatorial problems. Basically, we are considering \NP-hard problems. For those problemsrnwerncan not find a polynomial time algorithm. Several algorithmic approaches already exist which deal with this dilemma. Amongrnthemrnwe find (randomized) approximation algorithms and heuristics. Even though in practice they often work in reasonable time they usually do not return anrnoptimal solution. If we constrain optimality then there are only two methods which suffice for this purpose: exponential time algorithms andrnparameterized algorithms. In the first approach we seek to design algorithms consuming exponentially many steps who are more clever than some trivialrnalgorithm (whornsimply enumerates all solution candidates).rnTypically, the naive enumerative approach yields an algorithm with run time $\Oh^*(2^n)$. So, the general task is to construct algorithms obeying a run time of rnthe form $\Oh^*(c^n)$ where $c<2$.rn The second approach considers an additional parameter $k$ besides the input size $n$. This parameter shouldrnprovide more information about the problem and cover a typical characteristic. The standard parameterization is to see $k$ as an upper (lower, resp.)rnbound on the solution size in case of a minimization (maximization, resp.) problem. Then a parameterized algorithm should solve the problem in time $f(k)\cdot n^\beta$rnwhere $\beta$ is a constant and $f$ is independent of $n$. In principle this method aims to restrict the combinatorial difficulty of the problem tornthe parameter $k$ (if possible). The basic hypothesis is that $k$ is small with respect to the overall input size.rnIn both fields a frequent standard technique is the design of branching algorithms. These algorithms solve the problem by traversing the solutionrnspace in a clever way. They frequently select an entity of the input and create two new subproblems, one where this entity is considered as part ofrnthernfuture solution and another one where it is excluded from it. Then in both cases by fixing this entity possibly other entities will be fixed. If so then therntraversedrnnumber of possible solution is smaller than the whole solution space. The visited solutions can be arranged like a search tree. To estimate thernrun time of such algorithms there is need for a method to obtain tight upper bounds on the size of the search trees. In the field of exponential timernalgorithms a powerful technique called Measure&Conquer has been developed for this purpose. It has been applied successfully to manyrnproblems, especially to problems where other algorithmic attacks could not break the trivial run time upper bound. rnOn the other hand in the field of parameterized algorithms Measure&Conquer is almost not known. This piece of work will presentrnexamples where this technique can be used in this field. It also will point out what differences have to be made in order to successfully applyrnthe technique. Further, exponential time algorithms for hard problems where Measure&Conquer is applied are presented. Another aspect is thatrna formalization (and generalization) of the notion of a search tree is given. It is shown that for certain problems such a formalization is extremely useful.rn

We are living in a connected world, surrounded by interwoven technical systems. Since they pervade more and more aspects of our everyday lives, a thorough understanding of the structure and dynamics of these systems is becoming increasingly important. However - rather than being blueprinted and constructed at the drawing board - many technical infrastructures like for example the Internet's global router network, the World Wide Web, large scale Peer-to-Peer systems or the power grid - evolve in a distributed fashion, beyond the control of a central instance and influenced by various surrounding conditions and interdependencies. Hence, due to this increase in complexity, making statements about the structure and behavior of tomorrow's networked systems is becoming increasingly complicated. A number of failures has shown that complex structures can emerge unintentionally that resemble those which can be observed in biological, physical and social systems. In this dissertation, we investigate how such complex phenomena can be controlled and actively used. For this, we review methodologies stemming from the field of random and complex networks, which are being used for the study of natural, social and technical systems, thus delivering insights into their structure and dynamics. A particularly interesting finding is the fact that the efficiency, dependability and adaptivity of natural systems can be related to rather simple local interactions between a large number of elements. We review a number of interesting findings about the formation of complex structures and collective dynamics and investigate how these are applicable in the design and operation of large scale networked computing systems. A particular focus of this dissertation are applications of principles and methods stemming from the study of complex networks in distributed computing systems that are based on overlay networks. Here we argue how the fact that the (virtual) connectivity in such systems is alterable and widely independent from physical limitations facilitates a design that is based on analogies between complex network structures and phenomena studied in statistical physics. Based on results about the properties of scale-free networks, we present a simple membership protocol by which scale-free overlay networks with adjustable degree distribution exponent can be created in a distributed fashion. With this protocol we further exemplify how phase transition phenomena - as occurring frequently in the domain of statistical physics - can actively be used to quickly adapt macroscopic statistical network parameters which are known to massively influence the stability and performance of networked systems. In the case considered in this dissertation, the adaptation of the degree distribution exponent of a random, scale-free overlay allows - within critical regions - a change of relevant structural and dynamical properties. As such, the proposed scheme allows to make sound statements about the relation between the local behavior of individual nodes and large scale properties of the resulting complex network structures. For systems in which the degree distribution exponent cannot easily be derived for example from local protocol parameters, we further present a distributed, probabilistic mechanism which can be used to monitor a network's degree distribution exponent and thus to reason about important structural qualities. Finally, the dissertation shifts its focus towards the study of complex, non-linear dynamics in networked systems. We consider a message-based protocol which - based on the Kuramoto model for coupled oscillators - achieves a stable, global synchronization of periodic heartbeat events. The protocol's performance and stability is evaluated in different network topologies. We further argue that - based on existing findings about the interrelation between spectral network properties and the dynamics of coupled oscillators - the proposed protocol allows to monitor structural properties of networked computing systems. An important aspect of this dissertation is its interdisciplinary approach towards a sensible and constructive handling of complex structures and collective dynamics in networked systems. The associated investigation of distributed systems from the perspective of non-linear dynamics and statistical physics highlights interesting parallels both to biological and physical systems. This foreshadows systems whose structures and dynamics can be analyzed and understood in the conceptual frameworks of statistical physics and complex systems.