Classifier ensembles are mainly used for classification problems exhibiting complex data characteristics, e.g., a high class imbalance. They consist of a set of classifiers and a decision fusion method that combines the predictions of the classifiers. However, creating an accurate ensemble is a challenging optimization problem, since the individual classifiers and the decision fusion method must be selected from a huge search space that consists of numerous classification and fusion algorithms. Thereby, the selected classifiers have to be both accurate and diverse, so that they complement each other and make correct predictions on different data subsets. In addition, the selected decision fusion method has to accurately combine the predictions of individual classifiers into a consensus decision. Current literature focuses on AutoML approaches that reduce the complexity of the optimization problem by omitting the optimization of the classifier diversity and of the decision fusion. In this paper, we propose Auto-CEn as a novel ensemble approach based on AutoML that efficiently solves the optimization problem by selecting a set of accurate and diverse classifiers for the ensemble and by optimizing the decision fusion. In our extensive evaluation on 20 real-world datasets, we show that Auto-CEn outperforms several state-of-the-art baselines and that this improvement is mainly attributed to its novel contributions of diversity-based classifier selection and decision fusion optimization.

The increasing connectivity and data exchange between vehicles and the cloud have led to growing privacy concerns. To keep on gaining product insights through data collection while guaranteeing privacy protection, an anonymization-by-design approach should be used. A rising number of anonymization methods, not limited to the automotive domain, can be found in the literature and practice. The developers need support to select the suitable anonymization technique. To this end, we make the following two contributions: 1) We apply our knowledge from the automotive domain to outline the usage of qualitative metrics for anonymization techniques assessment; 2) We introduce HySAAD, a hybrid selection approach for anonymization by design that leverages this groundwork by recommending appropriate anonymization techniques for each mobile data analytics use case based on both, qualitative (i.e., "soft") metrics and quantitative (i.e., "hard") metrics. Using a real-world use case from the automotive, we demonstrate the applicability and effectiveness of HySAAD.

The extensive use of onboard sensors in smart cars enables the collection, processing, and dissemination of large amounts of mobile data containing information about the vehicle, its driver, and even bystanders. Despite the undoubted benefits of such smart cars, this leads to significant privacy concerns. Due to their inherent mobility, the situation of smart cars changes frequently, and with it, the appropriate measures to counteract the exposure of private data. However, data management in such vehicles lacks sufficient support for this privacy dynamism. We therefore introduce PaDS, a framework for Privacy adaptive Data Stream. The focus of this paper is to enable adaptive data processing within the vehicle data stream. With PaDS, Privacy-Enhancing Technologies can be deployed dynamically in the data pipeline of a smart car according to the current situation without user intervention. With a comparison of state-of-the-art approaches, we demonstrate that our solution is very efficient as it does not require a complete restart of the data pipeline. Moreover, compared to a static approach, PaDS causes only minimal overhead despite its dynamic adaptation of the data pipeline to react to changing privacy requirements. This renders PaDS an effective privacy solution for smart cars.

Efficient clustering algorithms, such as k -Means, are often used in practice because they scale well for large datasets. However, they are only able to detect simple data characteristics. Ensemble clustering can overcome this limitation by combining multiple results of efficient algorithms. However, analysts face several challenges when applying ensemble clustering, i. e., analysts struggle to (a) efficiently generate an ensemble and (b) combine the ensemble using a suitable consensus function with a corresponding hyperparameter setting. In this paper, we propose EffEns, an efficient ensemble clustering approach to address these challenges. Our approach relies on meta-learning to learn about dataset characteristics and the correlation between generated base clusterings and the performance of consensus functions. We apply the learned knowledge to generate appropriate ensembles and select a suitable consensus function to combine their results. Further, we use a state-of-the-art optimization technique to tune the hyperparameters of the selected consensus function. Our comprehensive evaluation on synthetic and real-world datasets demonstrates that EffEns significantly outperforms state-of-the-art approaches w.r.t. accuracy and runtime.

Combining two or more classifiers into an ensemble and fusing the individual classifier decisions to a consensus decision can improve the accuracy for a classification problem. The classification improvement of the fusion result depends on numerous factors, such as the data set, the combination scenario, the decision fusion algorithm, as well as the prediction accuracies and diversity of the multiple classifiers to be combined. Due to these factors, the best decision fusion algorithm for a given decision fusion problem cannot be generally determined in advance. In order to support the user in combining classifiers and to achieve the best possible fusion result, we propose the PUSION (Python Universal fuSION) framework, a novel generic and automated framework for decision fusion of classifiers. The framework includes 14 decision fusion algorithms and covers a total of eight different combination scenarios for both multi-class and multi-label classification problems. The introduced concept of AutoFusion detects the combination scenario for a given use case, automatically selects the applicable decision fusion algorithms and returns the decision fusion algorithm that leads to the best fusion result. The framework is evaluated with two real-world case studies in the field of fault diagnosis. In both case studies, the consensus decision of multiple classifiers and heterogeneous fault diagnosis methods significantly increased the overall classification accuracy. Our evaluation results show that our framework is of practical relevance and reliably finds the best performing decision fusion algorithm for a given combination task.

Analysts often struggle with the combined algorithm selection and hyperparameter optimization problem, a.k.a. CASH problem in literature. Typically, they execute several algorithms with varying hyperparameter settings to find configurations that show valuable results. Efficiently finding these configurations is a major challenge. In clustering analyses, analysts face the additional challenge to select a cluster validity index that allows them to evaluate clustering results in a purely unsupervised fashion. Many different cluster validity indices exist and each one has its benefits depending on the dataset characteristics. While experienced analysts might address these challenges using their domain knowledge and experience, especially novice analysts struggle with them. In this paper, we propose a new meta-learning approach to address these challenges. Our approach uses knowledge from past clustering evaluations to apply strategies that experienced analysts would exploit. In particular, we use meta-learning to (a) select a suitable clustering validity index, (b) efficiently select well-performing clustering algorithm and hyperparameter configurations, and (c) reduce the search space to suitable clustering algorithms. In the evaluation, we show that our approach significantly outperforms state-of-the-art approaches regarding accuracy and runtime.

Abstract Real-world data of multi-class classification tasks often show complex data characteristics that lead to a reduced classification performance. Major analytical challenges are a high degree of multi-class imbalance within data and a heterogeneous feature space, which increases the number and complexity of class patterns. Existing solutions to classification or data pre-processing only address one of these two challenges in isolation. We propose a novel classification approach that explicitly addresses both challenges of multi-class imbalance and heterogeneous feature space together. As main contribution, this approach exploits domain knowledge in terms of a taxonomy to systematically prepare the training data. Based on an experimental evaluation on both real-world data and several synthetically generated data sets, we show that our approach outperforms any other classification technique in terms of accuracy. Furthermore, it entails considerable practical benefits in real-world use cases, e.g., it reduces rework required in the area of product quality control.

The adoption of machine learning (ML) in organizations is characterized by the use of multiple ML software components. Citizen data scientists face practical requirements when building ML systems, which go beyond the known challenges of ML, e.g., data engineering or parameter optimization. They are expected to quickly identify ML system options that strike a suitable trade-off across multiple performance criteria. These options also need to be understandable for non-technical users. Addressing these practical requirements represents a problem for citizen data scientists with limited ML experience. This calls for a method to help them identify suitable ML software combinations. Related work, e.g., AutoML systems, are not responsive enough or cannot balance different performance criteria. In this paper, we introduce AssistML, a novel concept to recommend ML solutions, i.e., software systems with ML models, for predictive use cases. AssistML uses metadata of existing ML solutions to quickly identify and explain options for a new use case. We implement the approach and evaluate it with two exemplary use cases. Results show that AssistML proposes ML solutions that are in line with users' performance preferences in seconds.

Data has never been as significant as it is today. It can be acquired virtually at will on any subject. Yet, this poses new challenges towards data management, especially in terms of storage (data is not consumed during processing, i. e., the data volume keeps growing), flexibility (new applications emerge), and operability (analysts are no IT experts). The goal has to be a demand-driven data provisioning, i. e., the right data must be available in the right form at the right time. Therefore, we introduce a tailorable data preparation zone for Data Lakes called BARENTS. It enables users to model in an ontology how to derive information from data and assign the information to use cases. The data is automatically processed based on this model and the refined data is made available to the appropriate use cases. Here, we focus on a resource-efficient data management strategy. BARENTS can be embedded seamlessly into established Big Data infrastructures, e. g., Data Lakes.

Classification techniques are increasingly adopted for quality control in manufacturing, e.g., to help domain experts identify the cause of quality issues of defective products. However, real-world data often imply a set of analytical challenges, which lead to a reduced classification performance. Major challenges are a high degree of multi-class imbalance within data and a heterogeneous feature space that arises from the variety of underlying products. This paper considers such a challenging use case in the area of End-of-Line testing, i.e., the final functional test of complex products. Existing solutions to classification or data pre-processing only address individual analytical challenges in isolation. We propose a novel classification system that explicitly addresses both challenges of multi-class imbalance and a heterogeneous feature space together. As main contribution, this system exploits domain knowledge to systematically prepare the training data. Based on an experimental evaluation on real-world data, we show that our classification system outperforms any other classification technique in terms of accuracy. Furthermore, we can reduce the amount of rework required to solve a quality issue of a product.

Machine learning approaches may support various use cases in the manufacturing industry. However, these approaches often do not address the inherent characteristics of the real manufacturing data at hand. In fact, real data impose analytical challenges that have a strong influence on the performance and suitability of machine learning methods. This paper considers such a challenging use case in the area of End-of-Line testing, i.e., the final functional check of complex products after the whole assembly line. Here, classification approaches may be used to support quality engineers in identifying faulty components of defective products. For this, we discuss relevant data sources and their characteristics, and we derive the resulting analytical challenges. We have identified a set of sophisticated data-driven methods that may be suitable to our use case at first glance, e.g., methods based on ensemble learning or sampling. The major contribution of this paper is a thorough comparative study of these methods to identify whether they are able to cope with the analytical challenges. This comprises the discussion of both fundamental theoretical aspects and major results of detailed experiments we have performed on the real data of our use case.

\n The Internet of Things is applied in many domains and collects vast\n amounts of data. This data provides access to a lot of knowledge\n when analyzed comprehensively. However, advanced analysis techniques\n such as predictive or prescriptive analytics require access to both,\n history data, i.e., long-term persisted data, and real-time data as\n well as a joint view on both types of data. State-of-the-art hybrid\n processing architectures for big data - namely, the Lambda and the\n Kappa Architecture - support the processing of history data and\n real-time data. However, they lack of a tight coupling of the two\n processing modes. That is, the user has to do a lot of work manually\n in order to enable a comprehensive analysis of the data. For\n instance, the user has to combine the results of both processing\n modes or apply knowledge from one processing mode to the other.\n Therefore, we introduce a novel hybrid processing architecture for\n big data, called BRAID. BRAID intertwines the processing of history\n data and real-time data by adding communication channels between the\n batch engine and the stream engine. This enables to carry out\n comprehensive analyses automatically at a reasonable overhead.\n

On-the-fly data integration, i.e. at query time, happens mostly in tightly coupled, homogeneous environments where the partitioning of the data can be controlled or is known in advance. During the process of data fusion, the information is homogenized and data inconsistencies are hidden from the application. Beyond this, we propose in this paper the Nexus metadata model and a processing approach that support on-the-fly data integration in a loosely coupled federation of autonomous data providers, thereby advancing the status quo in terms of flexibility and expressive power. It is able to represent data and schema inconsistencies like multi-valued attributes and multi-typed objects. In an open environment, this best suites the application needs where the data processing infrastructure is not able to decide which attribute value is correct. The Nexus metadata model provides the foundation for integration schemata that are specific to a given application domain. The corresponding processing model provides four complementary query semantics in order to account for the subtleties of multi-valued and missing attributes. In this paper we show that this query semantics is sound, easy to implement, and it builds upon existing query processing techniques. Thus the Nexus metadata model provides a unique level of flexibility for on-the-fly data integration.

Industrial enterprise data present classication problems which are different from those problems typically discussed in the scientic community { with larger amounts of classes and with domain-specic, often unstructured data. We ad- dress one such problem through an analytics environment which makes use of domain-specic knowledge. Companies are beginning to use analytics on large amounts of text data which they have access to, but in day-to-day business, man- ual effort is still the dominant method for processing un- structured data. In the face of ever larger amounts of data, faster innovation cycles and higher product customization, human experts need to be supported in their work through data analytics. In cooperation with a large automotive man- ufacturer, we have developed a use case in the area of quality management for supporting human labor through text ana- lytics: When processing damaged car parts for quality im- provement and warranty handling, quality experts have to read text reports and assign error codes to damaged parts. We design and implement a system to recommend likely er- ror codes based on the automatic recognition of error men- tions in textual quality reports. In our prototypical imple- mentation, we test several methods forltering out accurate recommendations for error codes and develop further direc- tions for applying this method to a competitive business intelligence use case.

Smart devices became Marc Weiser's Computer of the 21st Century. Due to their versatility a lot of private data enriched by context data are stored on them. Even the health industry utilizes smart devices as portable health monitors and enablers for telediagnosis. So they represent a severe risk for information security. Yet the platform providers' countermeasures to these threats are by no means sufficient. In this paper we describe how information security can be improved. Therefore, we postulate requirements towards a secure handling of data. Based on this requirements specification, we introduce a secure data container as an extension for the Privacy Management Platform. Since a complete isolation of an app is usually not practicable, our approach also provides secure data sharing features. Finally, we evaluate our approach from a technical point of view as well as a security point of view and show its applicability in an eHealth scenario.

Today, a multitude of highly-connected applications and information systems hold, consume and produce huge amounts of heterogeneous data. The overall amount of data is even expected to dramatically increase in the future. In order to conduct, e.g., data analysis, visualizations or other value-adding scenarios, it is necessary to integrate specific, relevant parts of data into a common source. Due to oftentimes changing environments and dynamic requests, this integration has to support ad-hoc and flexible data processing capabilities. Furthermore, an iterative and explorative trial-and-error integration based on different data sources has to be possible. To cope with these requirements, several data mashup platforms have been developed in the past. However, existing solutions are mostly non-extensible, monolithic systems or applications with many limitations regarding the mentioned requirements. In this paper, we introduce an approach that copes with these issues (i) by the introduction of patterns to enable decoupling from implementation details, (ii) by a cloud-ready approach to enable availability and scalability, and (iii) by a high degree of flexibility and extensibility that enables the integration of heterogeneous data as well as dynamic (un-)tethering of data sources. We evaluate our approach using runtime measurements of our prototypical implementation.

Nowadays, mobile platform vendors have to concern themselves increasingly about how to protect their users' privacy. As Google is less restrictive than their competitors regarding their terms of use for app developers, it is hardly surprising that malware spreads even in Google Play. To make matters worse, in Android every user is responsible for his or her private data and s/he is frequently overwhelmed with this burden because of the fragile Android permission mechanism. Thus, the calls for a customizable, fine-grained, context-based, crash-proof, and intuitive privacy management system are growing louder. To cope with these requests, we introduce the Privacy Management Platform (PMP) and we discuss three alternative implementation strategies for such a system.

Scientific workflows often have to process huge data sets in a multiplicity of data formats. For that purpose, they typically embed complex data provisioning tasks that transform these heterogeneous data into formats the underlying tools or services can handle. This results in an increased complexity of workflow design. As scientists typically design their scientific workflows on their own, this complexity hinders them to concentrate on their core issue, namely the experiments, analyses, or simulations they conduct. In this paper, we present the core idea of a pattern-based approach to alleviate the design of scientific workflows. This approach is particularly targeted at the needs of scientists. We exemplify and assess the pattern-based design approach by applying it to a complex scientific workflow realizing a real-world simulation of structure changes in bones.

The consolidation of multiple tenants onto a single RDBMS instance turned out to be benefical with respect to resource utilization and scalability. The consolidation implies that multiple tenants share the physical resources available for the RDBMS instance. If the available resources tend to get insufficient to meet the SLAs agreed with the tenants, migration of a tenant's database from one RDBMS instance to another is compelling. Highly available services demand for live migration techniques that come with minimal service interruption and low performance impact.

A steadily growing number of people using location based services (LBS) inflict massive query loads on the data tier of an LBS. As such queries usually possess considerable overlap, multiple cache nodes collaborating in a distributed spatial cache can provide scalable access to frequently used data. To preserve high throughput throughout the complete execution process, it is necessary to balance the accumulating load among the participating cache nodes. In this work, we identify three key-indicators to improve resource utilization during the load-balancing process: data skew, anticipated data access patterns and dynamic load peaks. For this reason, we introduce a comprehensive mathematical model to express the key-indicators as probability distribution functions. We fuse the different key-indicators into a single holistic distribution model. In the course of this, we devise a methodology from our holistic distribution model towards a distributed spatial cache offering improved load-balancing facilities.

The still rising popularity of modern mobile phones results in an increased demand for manifold applications for these devices. As Android OS supports the development and usage of third-party software greatly, there are more and more developers for this platform. However, many of those applications handle private data grossly negligent which immediately leads to serious privacy concerns. To make matters worse, the current Android permission rules are much too coarse and incomprehensible from the average user's perspective. But even if s/he understands the meaning of the permissions, s/he must either accept all of them or waive the application. Therefore we review concepts and approaches towards effective privacy management for mobile platforms. All this is discussed based on the prevailing key players in the mobile market, namely Apple, RIM, Microsoft and Google. As this work has been initiated by Google we mainly concentrated on Android-based concepts towards customizable privacy management approaches. As a result of our review and taking into account current initiatives and trends in the market, we come up with a novel approach, an implementation architecture and a prototype.

Location-based services (LBS) have gained enormous popularity, which imposes increasing query loads at the data tier of an LBS. Yet, the data access patterns of LBS typically possess high temporal and spatial locality. Therefore, a dedicated spatial cache which provides efficient access to the data currently needed may considerably reduce this load. To ensure high throughput, multiple cache nodes can collaborate in a distributed spatial cache overlay, which balances load among the nodes. However, load-balancing is a non-trivial task in this context, as load spreads unevenly in space and varies notably over time. This requires constant readjustment to shifting hot spots. We present an elastic load-balancing mechanism between cache nodes that is based on the physical model of a particle-spring system. Using spring contraction, nodes instantly form processing clusters in regions with high load and thus can easily work off accumulating queries. Our evaluation shows that our approach quickly adapts to rapidly changing hot spots and thereby ensures high throughput throughout the entire execution.

The W3C Resource Description Framework (RDF) is gaining popularity for its ability to manage semi-structured data without a predefined database schema. So far, most RDF query processors have concentrated on finding complex graph patterns in RDF, which typically involves a high number of joins. This works very well to query resources by the relations between them. Yet, obtaining a record-like view on the attributes of resources, as natively supported by RDBMS, imposes unnecessary performance burdens, as the individual attributes must be joined to assemble the final result records. We present an approach to retrieve the attributes of resources efficiently. We first determine the resources in question and then retrieve all their attributes efficiently at once, exploiting contiguous storage in RDF indexes. In addition, we present an index structure which is specifically designed for RDF attribute retrieval. Our measurements show that our approach is clearly superior for larger numbers of attributes.

Software as a Service (SaaS) facilitates acquiring a huge number of small tenants by providing low service fees. To achieve low service fees, it is essential to reduce costs per tenant. For this, consolidating multiple tenants onto a single relational schema instance turned out beneficial because of low overheads per tenant and scalable manageability. This approach implements data isolation between tenants, per-tenant schema extension and further tenant-centric data management features in application logic. This is complex, disables some optimization opportunities in the RDBMS and represents a conceptual misstep with Separation of Concerns in mind.

Nowadays, many applications processes stream-based data, such as financial market analysis, network intrusion detection, or visualization applications. To process stream-based data in an application-independent manner, distributed stream processing systems emerged. They typically translate a query to an operator graph, place the operators to stream processing nodes, and execute them to process the streamed data. The operator placement is crucial in such systems, as it deeply influences query execution. Often, different stream-based applications require dedicated placement of query graphs according to their specific objectives, e.g. bandwidth not less than 500 MBit/s and costs not more that 1 cost unit. This fact constraints operator placement. Existing approaches do not take into account application-specific objectives, thus not reflecting application-specific placement decisions. As objectives might conflict among each other, operator placement is subject to delicate trade-offs, such as bandwidth maximization is more important than cost reduction. Thus, the challenge is to find a solution which considers the application-specific objectives and their trade-offs.

The mobile data management community is experiencing a rapid evolutionary change due to the worldwide diffusion of always-on mobile devices and to the increased popularity of location and context-aware mobile applications. Accordingly to recent studies, in two years from now one fourth of the total mobile data will come from audio and video streaming and nearly all the rest from other Internet services. A large part of the increase in mobile data will come from cloud computing applications that are massively used for storing personal data, for sharing data, as well as for utility software (such as maps) and productivity tools. Social networking will strongly influence the way mobile users choose, share and use content from mobile devices. On the other side mobile devices are changing the way social networks have been used till now introducing geo-tagging, location sharing, and many innovative location based services.

Mobile devices have become general-purpose computers that are equipped with sensors, constantly access the internet, and almost always accompany the user. Consequently, devices manage many different kinds of data about the user's life and context. There is considerable overlap in this data, as different applications handle similar data domains. Applications often keep this data in separated data silos. Web applications, which manage large amounts of personal data, hardly share this data with other applications at all. This lack of interoperability creates redundancy and impacts usability of mobile devices. We present a data management architecture for mobile devices to support interoperability between applications, devices and web applications at the data management level. We propose a central on-device repository for applications to share resource and context data in an integrated, extensible data model which uses semantic web technologies and supports location data. A web browser interface shares data with web applications, as controlled by a general security model.

Location-based services (LBS) have gained tremendous popularity with millions of simultaneous users daily. LBS handle very large data volumes and face enormous query loads. Both the data and the queries possess high locality: spatial data is distributed very unevenly around the globe, query load is different throughout the day, and users often search for similar things in the same places. This causes high load peaks at the data tier of LBS, which may seriously degrade performance. To cope with these load peaks, we present DiSCO, a distributed semantic cache overlay for LBS. DiSCO exploits the spatial, temporal and semantic locality in the queries of LBS and distributes frequently accessed data over many nodes. Based on the Content-Addressable Network (CAN) peer-to-peer approach, DiSCO achieves high scalability by partitioning data using spatial proximity. Our evaluation shows that DiSCO significantly reduces queries to the underlying data tier.

Online Analytical Processing (OLAP) is a popular technique for explorative data analysis. Usually, a fixed set of dimensions (such as time, place, etc.) is used to explore and analyze various subsets of a given, multi-dimensional data set. These subsets are selected by constraining one or several of the dimensions, for instance, showing sales only in a given year and geographical location. Still, such aggregates are often not enough. Important information can only be discovered by combining several dimensions in a multidimensional analysis. Most existing approaches allow to add new dimensions either statically or dynamically. These approaches support, however, only the creation of global dimensions that are not interactive for the user running the report. Furthermore, they are mostly restricted to data clustering and the resulting dimensions cannot be interactively refined.

Semantic Web technologies, most notably RDF, are well-suited to cope with typical challenges in spatial data management including analyzing complex relations between entities, integrating heterogeneous data sources and exploiting poorly structured data, e.g., from web communities. Also, RDF can easily represent spatial relationships, as long as the location information is symbolic, i.e., represented by places that have a name. What is widely missing is support for geographic and geometric information, such as coordinates or spatial polygons, which is needed in many applications that deal with sensor data or map data. This calls for efficient data management systems which are capable of querying large amounts of RDF data and support spatial query predicates. We present a native RDF triple store implementation with deeply integrated spatial query functionality. We model spatial features in RDF as literals of a complex geometry type and express spatial predicates as SPARQL filter functions on this type. This makes it possible to use W3C's standardized SPARQL query language as-is, i.e., without any modifications or extensions for spatial queries. We evaluate the characteristics of our system on very large data volumes.

With the increasing use of sensor technology, the compression of sensor data streams is getting more and more important to reduce both the costs of further processing as well as the data volume for persistent storage. A popular method for sensor data compression is to smooth the original measurement curve by an approximated curve, which is bounded by a given maximum error value. Measurement values from positioning systems like GPS are an interesting special case, because they consist of two spatial and one temporal dimension. Therefore various standard techniques for approximation calculations like regression or line simplification algorithms cannot be directly applied. In this paper, we portray our stream data management system NexusDS and an operator for compressing sensor data. For the operator, we implemented various compression algorithms for position data streams. We present the required adaptations and the different characteristics of the compression algorithms as well as the results of our evaluation experiments, and compare them with a map matching approach, specifically developed for position data.

Ad-hoc smart spaces aim at resource-rich mobile devices sharing resource and context data with others near them spontaneously. Thus, a device may, e.g., obtain a more complete context model by utilizing sensor data of its neighbors via wireless communication, such as Bluetooth. The highly dynamic device neighborhood challenges resource discovery, as the devices have to organize themselves autonomously. This paper evaluates different resource discovery protocols in Bluetooth-based ad-hoc smart spaces. We simulate the protocols in different scenarios taking into account the scatternet structure of the network. We suggest request flooding for small settings and random replication for medium to large spaces.

Techniques for efficient and distributed processing of huge, unbound data streams have made some impact in the database community. Sensors and data sources, such as position data of moving objects, continuously produce data that is consumed, e.g., by location-aware applications. Depending on the domain of interest, e.g. visualization, the processing of such data often depends on domain-specific functionality. This functionality is specified in terms of dedicated operators that may require specialized hardware, e.g. GPUs. This creates a strong dependency which a data stream processing system must consider when deploying such operators. Many data stream processing systems have been presented so far. However, these systems assume homogeneous computing nodes, do not consider operator deployment constraints, and are not designed to address domain-specific needs.In this paper, we identify necessary features that a flexible and extensible middleware for distributed stream processing of context data must satisfy. We present NexusDS, our approach to achieve these requirements. In NexusDS, data processing is specified by orchestrating data flow graphs, which are modeled as processing pipelines of predefined and general operators as well as custom-built and domain-specific ones. We focus on easy extensibility and support for domain-specific operators and services that may even utilize specific hardware available on dedicated computing nodes.

Over the last years, data management products as well as workflow products have established themselves as indispensable building blocks for advanced IT systems in almost all application areas. Recently, many vendors have created innovative product extensions that combine service-oriented frameworks with powerful workflow and data management capabilities. In this paper, we discuss several workflow products from different vendors with a specific focus on their SQL support. We provide a comparison based on a set of important data management patterns and illustrate the characteristics of various approaches by means of a running example.

Nowadays XML-based data integration systems are accepted as data service providers on the web. In order to make such a data integration system fully equipped with data manipulation capabilities, programming frameworks which support update at the integration level are being developed. When the user is permitted to submit updates, it is necessary to establish the best possible data consistency in the whole data integration system. To that extend, we present an approach based on an XQuery trigger service. We define an XQuery trigger model together with its semantics. We report on the integration of the XQuery trigger service into the overall architecture and discuss details of the execution model. Experiments show that data consistency is enforced easily, efficiently and conveniently at the global level.

We present the design and implementation of a component for the preprocessing of position data taken from moving objects. The movement of mobile objects is represented by piecewise functions over time that approximate the real object movement and significantly reduce the initial data volume such that effcient storage and analysis of object trajectories can be achieved. The maximal acceptable deviation - an input parameter of our algorithms - of the approximations also includes the uncertainty of the position sensor measurements. We analyze and compare five different lossy preprocessing methods. Our results clearly indicate that even with simple approaches, a more than sufficient overall performance can be achieved.

This paper presents a new approach for the development of data-intensive web applications that depend on sophisticated application logic. E-Commerce web sites, on-line auction systems and large enterprise web portals fall into this category as they require comprehensive data access, data processing and data manipulation capabilities. However, existing methodologies mainly concentrate on modeling content, navigation and presentation aspects of read-only web sites. In our opinion these models are not sufficient to express complex operations that access or modify web application content. Therefore, we propose an additional Operation Model defining the operation logic of a web application. We show that based on this model a significant part of a web application's Operation Layer can be generated, still allowing the manual implementation of arbitrary additional functionality. We evaluate our approach and present experimental results based on a large example application for the area of innovation management.

In enterprises, information is encoded in documents and databases. Logically, the information in both worlds is tightly connected, however, on the system level there is usually a large gap. In this paper, we propose a framework that improves document retrieval by exploiting available enterprise databases. In particular, we use database information to model the context of documents and incorporate this context in our search framework. We present our framework architecture, its components and its major interfaces. The framework can be configured and enhanced at well-defined points and, hence, can easily be customized to other domains. We furthermore evaluate its core components. Our experiments show that the context-aware approach significantly improves the quality of search results

Information in enterprises comes in documents and data bases. From a semantic viewpoint, both kinds of information are usually tightly connected. In this paper, we propose to enhance common search-engines with contextual information retrieved from databases. We establish system requirements and anecdotally demonstrate how documents and database information can be represented as the nodes of a graph. Then, we give an example how we exploit this graph information for document retrieval.

Many decision support applications are built upon data mining and OLAP tools and allow users to answer information requests based on a data warehouse that is managed by a powerful DBMS. We focus on tools that generate sequences of SQL statements in order to produce the requested information. Our thorough analysis revealed that many sequences of queries that are generated by commercial tools are not very efficient. An optimized system architecture is suggested for these applications. The main component is a DSS optimizer that accepts previously generated sequences of queries and remodels them according to a set of optimization strategies, before they are executed by the underlying database system. The advantages of this extended architecture are discussed and a couple of appropriate optimization strategies are identified. Experimental results are given, showing that these strategies are appropriate to optimize query sequences of OLAP applications.

Today many applications routinely generate large quantities of data. The data often takes the form of (time) series, or more generally streams, i.e. an ordered sequence of records. Analysis of this data requires stream processing techniques which differ in significant ways from what current database analysis and query techniques have been optimized for. In this paper we present a new operator, called StreamJoin, that can efficiently be used to solve stream-related problems of various applications, such as universal quantification, pattern recognition and data mining. Contrary to other approaches. StreamJoin processing provides rapid response times, a non-blocking execution as well as economical resource utilization. Adaptability to different application scenarios is realized by means of parameters. In addition, the StreamJoin operator can be efficiently embedded into the database engine, thus implicitly using the optimization and parallelization capabilities for the benefit of the application. The paper focuses on the applicability of StreamJoin to integrate application semantics into the DBMS.

We describe the design and implementation of a CORBA query service targeted to access data that is defined by the EXPRESS data modeling language. EXPRESS is used primarily in engineering domains (like CAD/CAM and GIS) to describe mostly product model data (like parts explosion or product geometry). In order to bring query facilities for EXPRESS-based data to CORBA a number of design decisions have to be taken, although the CORBA query service is standardized by the OMG. Among the most important and performance-indicating decisions are the definition of an appropriate query language and the description of the query result data structures. We discuss solutions to these topics and report on the experiences gained in designing and implementing our first CORBA query service for EXPRESS-based data.

Intra/Internet technology has become a key issue in the development of modern systems. Nowadays, it is no longer sufficient to present static information sheets through the World Wide Web (WWW). Instead, we need interactive applications that may even compute complex results or process large data sets. In this paper, we describe a prototype based on Java and CORBA. Both represent modern concepts that have been developed to fulfill these requirements. Their combination results into the kind of data processing we want to apply to the WWW: (1) portable, powerful, structured and even reusable client programs instead of cryptic HTML scripts; (2) well-defined interfaces, and (3) efficient server processes separated from the WWW server and its CGI extensions. Communication is controlled by a fault-tolerant CORBA layer, which also enables server development using a different language than Java. Besides a discussion of CORBA and its data shipping capabilities, we take a closer look at Java and its runtime behavior, and we report on the experiences gathered with our prototype system and its testbed application. This system has also been used to gather experiences with and to influence the new language binding of the Standard Data Access Interface (SDAI) of the Standard for the Exchange of Product Data (STEP, ISO 10303) to Java.

Nowadays parallel object-relational DBMS are envisioned as the next great wave, but there is still a lack of efficient implementation concepts for some parts of the proposed functionality. Thus one of the current goals for parallel object-relational DBMS is to move towards higher performance. In this paper we develop a framework that allows to process user-defined functions with data parallelism. We will describe the class of partitionable functions that can be processed parallelly. We will also propose an extension which allows to speed up the processing of another large class of functions by means of parallel sorting. Functions that can be processed by means of our techniques are often used in decision support queries on large data volumes, for example. Hence a parallel execution is indispensable.

Modern document management systems administer an increasing amount of data. Unfortunately traditional database technology cannot satisfy the resulting performance requirements. We report on how parallel database technology can contribute to make large and scalable document management systems work. Our discussion clearly shows that simply substituting a sequential database system for a parallel one does not necessarily lead to scalability and high performance. In contrast, detailed knowledge of the application and document management system is necessary to decide on the optimal architecture of the parallel database system, its query optimization and parallelization strategies, as well as on its data fragmentation scheme and load balancing properties. All discussions are exemplified referring to an existing document management system developed at our university that is currently migrated from a sequential database system to a parallel one.

'Computer-Supported Cooperative Work' is a young research area considering applications with strong demands on database technology. Design applications need support for cooperation and some means for controlling their inherent dynamics. However, today's CAD systems consisting of a collection of diverse design tools typically do not support these requirements. Therefore, an encompassing processing model is needed that covers the overall design process in general as well as CAD-tool application in particular. The CONCORD model described in the paper, reflects the distinct properties of design process dynamics by distinguishing three levels of abstraction. The highest level supports application-specific cooperation control and design process administration, the second considers goal-oriented tool invocation and work-flow management while the third level provides tool processing of design data. To achieve level-spanning control, the authors rely on transactional facilities provided at the various system layers.>

A great variety of techniques has been developed to optimize and enhance query processing for relational, client/server, distribute~ parallel, and heterogeneous database systems (DBS).Baaed on that work and experience, we investigate how far those techniques are applicable to query processing in Knowledge Base Management Systems (KBMS).Our reference system is the KRISYS KBMS that consists of a knowledge-processing system at the client (clientbaaed processing) and a data-processing system at the server (database backend).We describe a unifying thrnework for query ,processing incorporating both processing systems (as realized in KRISYS).This allows to distribute and balance the amount of work done in the client and in the server.Based on an evaluation of that framework, severaf approaches to further enhance knowledge processing are reported.

An extension to SQL, called the SQL extended normal form (XNF), is discussed. It enhances relational technology by a composite object facility, which comprises not only extraction of composite objects from existing databases but also efficient navigation and manipulation facilities provided by an appropriate application programming interface. The language itself allows sharing of the database among normal form SQL applications and composite object applications. It provides proper subsetting of the database and subsequent structuring, exploiting subobject sharing and recursion, all based on its powerful composite object constructor concept, which is closed under the language operations. XNF is integrated into the relational framework, thus benefiting from the available technology such as relational engine and query optimization.>

PRIMA is a non-standard database system developed at the University Kaiserslautern. Its major purpose is the support of engineering design applications, such as VLSI design and software engineering. The applications require tailored application-dependent interfaces which, however, all share basic notions like that of a composite object. Hence, the approach of PRIMA is to offer an application-independent complex-object interface (the moleculeatom data model, shortly called MAD model) and to provide means to easily augment this interface by application-dependent functionality. In the following, we will concentrate on the MAD model and its implementation.