05 Fakultät Informatik, Elektrotechnik und Informationstechnik
Permanent URI for this collectionhttps://elib.uni-stuttgart.de/handle/11682/6
Browse
4 results
Search Results
Item Open Access System support for adaptive pervasive applications(2009) Handte, Marcus; Rothermel, Kurt (Prof. Dr. rer. nat. Dr. h. c.)Driven by the ongoing miniaturization of computer technology as well as the proliferation of wireless communication technology, Pervasive Computing envisions seamless and distraction-free task support by distributed applications that are executed on computers embedded in everyday objects. As such, this vision is equally appealing to the computer industry and the user. Induced by various factors such as invisible integration, user mobility and computer failures, the resulting computer systems are heterogeneous, highly dynamic and evolving. As a consequence, applications that are executed in these systems need to adapt continuously to their ever-changing execution environment. Without further precautions, the need for adaptation can complicate application development and utilization which hinders the realization of the basic vision. As solution to this dilemma, this dissertation describes the design of system software for Pervasive Computing that simplifies the development of adaptive applications. As opposed to shifting the responsibility for adapting an application to the user or the application developer, the system software introduces a component-based application model that can be configured and adapted automatically. To enable automation at the system level, the application developer specifies the dependencies on components and resources in an abstract manner using contracts. Upon application startup, the system uses the contractual descriptions to compute and execute valid configurations. At runtime, it detects changes to the configuration that require adaptation and it reconfigures the application. To compute valid configurations upon application startup, the dissertation identifies the requirements for configuration algorithms. Based on an analysis of the problem complexity, the dissertation classifies possible algorithmic solutions and it presents an integrated approach for configuration based on a parallel backtracking algorithm. Besides from scenario specific modifications, retrofitting the backtracking algorithm requires a problem mapping from configuration to constraint satisfaction which can be computed on-the-fly at runtime. The resulting approach for configuration is then extended to support the optimization of a cost function that captures the most relevant cost factors during adaptation. This enables the use of the approach for configuration upon startup and reconfiguration during runtime adaptation. As basis for the evaluation of the system software and the algorithm, the dissertation outlines a prototypical implementation. The prototypical implementation is used for a thorough evaluation of the presented concepts and algorithms by means of real world measurements and a number of simulations. The evaluation results suggest that the presented system software can indeed simplify the development of distributed applications that compensate the heterogeneity, dynamics and evolution of the underlying system. Furthermore, they indicate that the algorithm for configuration and the extensions for adaptation provide a sufficiently high performance in typical applications scenarios. Moreover, the results also suggest that they are preferable over of alternative solutions. To position the presented solution within the space of possible and existing solutions, the dissertation discusses major representatives of existing systems and it proposes a classification of the relevant aspects. The relevant aspects are the underlying conceptual model of the system and the distribution of the responsibility for configuration and adaptation. The classification underlines that in contrast to other solutions, the presented solution provides a higher degree of automation without relying on the availability of a powerful computer. Thus, it simplifies the task of the application developer without distracting the user while being applicable to a broader range of scenarios. After discussing the related approaches and clarifying similarities and differences, the dissertation concludes with a short summary and an outlook on future work.Item Open Access Optimization of query sequences(2009) Kraft, Tobias; Mitschang, Bernhard (Prof. Dr.-Ing. habil.)Query optimization is a well-known topic in database research since the 1970s. This thesis highlights a special area of query optimization that arises from new trends in the usage of databases. Whereas in the beginning databases were primarily used for transaction-oriented processing of operative data, today databases are also used to facilitate reporting and analysis on consolidated, historic data. For the latter, the data is loaded into a large data warehouse and afterwards it is being analyzed by the use of tools. The tools used to model the flows that extract the operative data from the source systems, transform these data and load it into the data warehouse as well as the tools that process the data stored in the data warehouse often generate sequences of SQL statements that break down a complex flow or request into a sequence of computational steps. The optimization of this kind of sequences with respect to runtime is the focus of this thesis. We propose a heuristic as well as a cost-based approach for this optimization problem. The cost-based approach is just an enhancement of the heuristic approach. It results from adding a cost estimation component to the optimizer architecture of the heuristic approach and by replacing the heuristic control strategy by a control strategy that considers cost estimates. Both approaches are rule-based approaches that rewrite a given sequence of SQL statements into a syntactically different but semantically equivalent sequence of SQL statements. Therefore, we specify a set of rewrite rules. For cost estimation, we employ the capabilities of the query optimizer of the underlying database management system (DBMS) which is responsible for the execution of the query sequences. To improve the quality of these cost estimates, we support the query optimizer of the underlying DBMS with statistics that we derive from histogram propagation. For this purpose, we need an interface that allows to access and manipulate statistics in the underlying DBMS. Since there exists no standardized interface for this purpose, we define our own DBMS-independent interface. For the heuristic approach as well as for the cost-based approach, we provide prototypic implementations in JAVA. Furthermore, we have implemented the DBMS-independent interface for the three commercial DBMSs IBM DB2, Oracle, and Microsoft SQL Server. We report on the results of experiments that we conducted with our prototypes and some sample sequences that we derived by using a commercial tool for online analytical processing (OLAP). They show the effectiveness of our optimization approach and they highlight the optimization potential that lies in rewriting sequences of SQL statements. Finally, we draw a conclusion and suggest some interesting points for future research.Item Open Access Fundamental storage mechanisms for location based services in mobile ad hoc networks(2009) Dudkowski, Dominique; Rothermel, Kurt (Prof. Dr. rer. nat. Dr. h. c.)The proliferation of mobile wireless communication technology has reached a considerable magnitude. As of 2009, a large fraction of the people in most industrial and emerging nations is equipped with mobile phones and other types of portable devices. Supported by trends in miniaturization and price decline of electronic components, devices become enhanced with localization technology, which delivers, via the Global Positioning System, the geographic position to the user. The combination of both trends enables location-based services, bringing information and services to users based on their whereabouts in the physical world, for instance, in the form of navigation systems, city information systems, and friend locators. A growing number of wireless communication technologies, such as Wireless Local Area Networks, Bluetooth, and ZigBee, enable mobile devices to communicate in a purely peer-to-peer fashion, thereby forming mobile ad-hoc networks. Together with localization technology, these communication technologies make it feasible, in principle, to implement distributed locationbased services without relying on any support by infrastructure components. However, the specific characteristics of mobile ad-hoc networks, especially the significant mobility of user devices and the highly dynamic topology of the network, make the implementation of locationbased services extremely challenging. Current research does not provide an adequate answer to how such services can be supported. Efficient, robust, and scalable fundamental mechanisms that allow for generic and accurate services are lacking. This dissertation presents a solution to the fundamental support of location-based services in mobile ad-hoc networks. A conceptual framework is outlined that implements mechanisms on the levels of routing, data storage, location updating, and query processing to support and demonstrate the feasibility of location-based services in mobile ad-hoc networks. The first contribution is the concept of location-centric storage and the implementation of robust routing and data storage mechanisms in accordance with this concept. This part of the framework provides a solution to the problems of data storage that stem from device mobility and dynamic network topology. The second contribution is a comprehensive set of algorithms for location updating and the processing of spatial queries, such as nearest neighbor queries. To address more realistic location-based application scenarios, we consider the inaccuracy of position information of objects in the physical world in these algorithms. Extensive analytical and numerical analyses show that the proposed framework of algorithms possesses the necessary performance characteristics to allow the deployment of location-based services in purely infrastructureless networks. A corollary from these results is that currently feasible location-based services in infrastructure-based networks may be extended to the infrastructureless case, opening up new business opportunities for service providers.Item Open Access Constraints and triggers to enhance XML-based data integration systems(2009) Lu, Jing; Mitschang, Bernhard (Prof. Dr.-Ing. habil.)XML is becoming one of the main technological integredients of the Internet. It is now accepted as the standard for information exchange. XML-based data integration system, which enables sharing and cooperation with legacy data sources, arises as a more and more important data service provider on the web. These services can provide the users with a uniform interface to a multitude of data sources such as relational databases, XML files, text files, delimited files, Excel files, etc. Users can thus focus on what they want, rather than think about how to obtain the answers. Therefore, users do not have to carry on the tedious tasks such as finding the relevant data sources, interacting with each data source in isolation using the local interface and combining data from multiple data sources. Users are always expecting better query performance and data consistency from the data integration systems. This work proposes an approach to support constraints and triggers in the XML-based data integration system in order to optimize queries and to enforce data consistency. Constraints and triggers have long been recognized to be useful in semantic query optimization and data consistency enforcement in relational databases. This work first gives an approach to use constraints from the heterogeneous data sources to semantically optimize queries submitted to the XML-based data integration system. Different constraints from the data sources are first integrated into a uniform constraint model. Then the constraints in the uniform constraint model are stored in the constraint repository. Traditional semantic query optimization techniques in the relational database are analyzed and three of them are reused and applied by the semantic query optimizer for XML-based data integration system. Among them are detection of empty results, join elimination and predicate elimination. Performance is analyzed according to the data source type and the data volume. The semantic query optimizer works best when the data sources are non-relational, the data volume is huge and the execution cost is expected to be high. In order to make the XML-based data integration system fully equipped with data manipulation capabilities, programming frameworks which support update at the integration level are being developed. This work discusses how to realize update in the XML-based data integration system under the Service Data Objects programming framework. When the user is permitted to submit updates, it is necessary to guarantee data integrity and enforce active business logics in the data integration system. This work presents an approach by which active rules including integrity constraints are enforced by XQuery triggers. An XQuery trigger model in conformance to XQuery update model proposed by W3C is defined. How to define active rules and integrity constraints by XQuery triggers is discussed. Triggers and constraints are stored in the trigger repository. The architecture supporting XQuery trigger service in the XML-based data integration system is proposed. Important components including event detection, trigger scheduling, condition evaluation, action firing and trigger termination are discussed. The whole XQuery trigger service architecture above a data integration system is implemented in BEA AquaLogic DataService Platform under the Service Data Objects programming framework. Experiments show active rules and integrity constraints are enforced easily, efficiently and conveniently at the global level. Constraints and triggers play an important role in XML-based data integration systems. Using constraints and triggers in the XML-based data integration system we can efficiently improve query performance and enforce data consistency.