A time series in its numeric form remains vulnerable to its nature, being highly dimensional. To be machine- and human- interpretable, a time series should be abstracted in a form to accommodate further computational tasks. As such, the form of abstraction may continue to be numerical (e.g., an aggregation feature) or lexical (i.e., text-based), which might increase the level of their interpretability (In machine learning and data mining communities, interpretability is the ability to explain or to present in understandable terms to a human [1].). In this paper, we propose the concept of extracting primitive lexical constructs, called shapeoids, which offers the ability to abstract a time series curve in natural language (i.e., interpretable), enabling tasks such as reasoning and information retrieval to occur on a later step. The definition of a shapeoid remains unchanged for any input time series. We provide formal evidence, using group theory, that our method will always enable us to extract shapeoids from a time series.

Time series representation and discretisation methods are susceptible to scaling over massive data streams. A recent approach for transferring time series data to the realm of symbols under primitives, named shapeoids has emerged in the area of data mining and pattern recognition. A shapeoid will characterise a subset of the time series curve in words from its morphology. Data processing frameworks are typical examples for running operations on top of fast unbounded data, with innate traits to enable other methods which are restricted to bounded data. Apache Beam is emerging with a unified programming model for streaming applications able to uniquely translate and run on multiple execution engines, saving development time to focus on other design decisions. We develop an application on Apache Beam which transfers the concept of shapeoids to a scenario in large-scale network flow monitoring infrastructure and evaluate it over two stream computing engines.

Ambient Assisted Living (AAL) is an application of Smart Environments, dealing with elderlies and their caregivers’ assistance in their daily life within their enhanced apartments. An AAL environment needs constant observation of the inhabitant’s activities to inform caregivers of critical situations respectively to react to them, such as the patient leaving the flat with the stove still on. Setting up an AAL environment is costly and complicated, as all sensors are tailored to the specific situation. Various industrial systems or research activities exist to monitor the environment and apply a rule-based inference to detect the multiple conditions as far as possible. There are, however, a range of standard day-to-day sensors, such as light switches, window sensors etc., which do not directly monitor patient conditions but allow for inference about a situation, e.g. whether a person has left the flat. We call this “lifted” contextual information. Also, there is much uncertainty in such environments, such as sensor malfunctions, power loss, or connectivity issues. Hence, a situation awareness system should freely combine and switch between combinations of sensors for identifying and verifying the current situation, respectively, inferences drawn from it. For example, confirm that the person has left the flat by checking for a webcam movement. This resembles our ability to use commonsense when we look at possible sensor readings on a dashboard. We make plain inferences based on a hypothesis on the given evidence. Such a system needs to make logical connections between different data and contribute to a derived information. We propose developing a logic-based system using the sensor events as evidence for a commonsense reasoning task.

The recognition and extraction of data-driven patterns is a challenging task. In vast amounts of data, suitable techniques should prepare data that match the user's intentions. For example, "an increasing daily price of a stock, is generally followed by a sudden fall due to the endorsed profit gains". Such a statement should effortlessly be transferred as input to a data mining task, without the need for an extensive training phase building an extremely complicated model and the usage of threshold-based approaches. We propose a method for extracting lexical representations directly from the raw data, enabling others not directly compatible with real-valued data. We provide evidence that our method is fast and accurate. We use as evaluation a preliminary step for a classification task compared to state-of-the-art classifiers applying it on a publicly available dataset.

In a world of uncertainty and incompleteness, one must "make sense" of found observations. Cyber-physical systems output large quantities of data, opening massive opportunities and challenges for scalable techniques to gain exciting insights. One intriguing challenge is the process of Sensor Data Sensemaking. The research presents an approach to handle this process by bringing together the strands of data and knowledge in a single architecture in an interpretable and expressive way. Differently from other works, the use of interpretable patterns from streaming data is in the spotlight. In addition, background knowledge over these patterns gasps the intention to give meaning to these patterns with several possible explanations. A hybrid implementation realises the approach following big data processing models.

With the always increasing availability of sensor devices, there is constant unseen monitoring of our environment. A physical activity has an impact on more sensor modalities than we could imagine. It is so vivid that distinctive patterns in the data look almost interpretable. Such knowledge, which is innate to humans, ought to be encoded and reason upon declaratively. We demonstrate the power of Markov Logic Networks for encoding uncertain knowledge to discover interesting situations from the observed evidence. We formally relate distinguishable patterns from the sensor data with knowledge about the environment and generate a rule basis for verifying and explaining occurred phenomena. We demonstrate an implementation on a real dataset and present our results.

Interpretation of the archaeological record is determined by the occurrence and distribution of finds and their (relative and absolute) distribution over space. Comparison and categorization of finds strongly depends on their characteristics and the significance assigned to them. By default this is a very erroneous process leading to debate about individual finds, their significance and of course their interpretation. In this paper we will discuss the (objective) impact of probability on interpretation of data by measuring and comparing the impact of probability on metrics such as centrality.

Methods from Social Network Analysis are frequently used to assess social relationships in prehistoric times. Archaeological data are primarily stochastic by nature and subject to interpretation - this is not in line with the methods of Social Network Analysis. This paper examines the concrete problems and suggests a method to deal with stochastic data in such analysis, thus allowing archaeological investigation. We suggest a "stability" measurement that allows assessing the "quality" of a Social Network Analysis - this is a novum for this type of analysis and allows archaeologists or sociologists to reason over the likelihood that a specific interpretation is correct.

—Applications need to be constantly re-developed for new devices and infrastructures, and to address new user needs. This leads to an increasing maintenance cost that only large-scale companies can afford. The problem with traditional Turing based programming models is that algorithms cannot be easily adjusted and thus bind the application to an environment. In this paper, we discuss how mathematical definitions can be used to not only describe algorithms, but specifically to allow their transformation and (re-)generation to principally address different infrastructures and requirements at considerably reduced effort.

—”The whole is more than the sum of its parts” (Aris-totle). Current imperative languages do not allow a program to be simply broken up (decomposition) or to merge several parts of a program, but demand appropriate knowledge and manual effort. The idea behind is to transfer methods from mathematical combinatorics to standard programming models to enable the distribution of a task across multiple heterogeneous resources. This approach allows distributed, heterogeneous resources to be treated as an integrated platform, with no hassle of adaptation for the developer. In this paper we propose and discuss a theoretical framework, with which the correctness of the code can be guaranteed with automated (de-)composition and adaptation. This will lay the groundwork for new programming methods that will allow code to be more fully understood, analysed and modified. This is relevant for all areas that develop and use software.

Over the last years, computer applications have moved away from the single-core, local execution that they were originally intended for. Modern applications have to run distributed, in parallel, on heterogeneous environments, share distributed data etc. Not surprisingly the standard ways of programming do not hold anymore and modern software engineers struggle with the growing complexity. New software engineering need to move away from the traditional execution models and reconsider the principles from ground up - focusing on the three primary “building criteria” of modern applications, namely to process information, to adhere to quality and performance criteria, and to follow business incentives. This paper shows that these three criteria are sufficient to specify the full functional behaviour of an application and thereby implicitly its potential instantiations. Since no single algorithm is bound to the specification, transformation between instances, and thus between different execution attributes is easily possible using mathematical transformations.

The world has moved from a single core, local application execution to distributed, parallel, social, heterogeneous etc. Standard ways of programming thus do not hold anymore and all software engineers struggle with the growing complexity that equally affects performance, maintainability and, worst of all, cost. The paper presents a concept arising from the EC Cloud Computing expert group between 2012 and 2016 as a means to move away from the hazards of traditional, Turing based computing concepts to an abstract and more natural way of programming. The concept builds up on three major building blocks, namely "Information", "Intention" and "Incentive" to define the behaviour of an application in a fashion that can abstract away from the actual way of executing an algorithm, thus allowing for a larger degree of adaptation, distribution and scale.

CACTOS offers Cloud developers, operators, and consultants a context-aware optimisation for private Clouds. It leads to better and more reliable user experience, by optimising the mapping of virtual to physical resources, considering application requirements and heterogeneity. The optimisation and simulation requires monitoring, and an integration for controlling and intercepting client requests.

IaaS Cloud systems enable the Cloud provider to overbook his data centre by selling more virtual resources than physical resources available. This approach works if on average the resource utilisation of a virtual machine is lower than the virtual machine boundaries. If this assumption is violated only locally, Cloud users will experience performance degradation and poor quality of service. This paper proposes the introduction of dynamic overbooking in the sense that the overbooking factors are not equal for all physical resources, but vary dynamically depending on the resource demands of the virtual resources they host. It allows new pricing models that are dependent on the overbooking a Cloud customer is willing to accept. Additionally, we discuss prerequisites for supporting its realisation in an OpenStack private Cloud, including a monitoring system, dedicated metrics to be monitored, as well as performance models that predict the performance degradation depending on the overbooking.

Model-drivenengineering promotes models and modeltrans- formations as the primary assets in software development. The models@run-time approach provides an abstract rep- resentation of a system at run-time, whereby changes in the model and the system are constantly reflected on each other. In this paper, we report on more than three years of experience with realising models@run-time in scalable cloud scenarios using a technology stack consisting of the Eclipse Modelling Framework (EMF) and Connected Data Objects(CDO).We establish requirements for the three roles domain-specific language (DSL) designer, developer, and operator, and compare them against the capabilities of EM- F/CDO. It turns out that this technology stack is well-suited for DSL designers, but less recommendable for developers and even less suited for operators. For these roles, we experi- enced a steep learning curve and several lacking features that hinder the implementation of models@run-time in scalable cloud scenarios. Performance experiences show limitations for write heavy scenarios with an increasing amount of total elements. While we do not discourage the use of EMF/CDO for such scenarios, we recommend that its adoption for sim- ilar use cases is carefully evaluated until this technology stack has realised our wish list of advanced features.

Infrastructure as a Service (IaaS) operators need to balance multiple adversarial goals, such as data center performance and energy efficiency. Automated resource management policies implemented in IaaS Cloud middleware allow the operators to automate trade-off decisions. Simulation-based analyses are viable means to validate that the utilized policies achieve the goals of the operator. For an IaaS operator to perform meaningful analyses, changes in the workload mix of active VMs need to be considered. Current Cloud simulation approaches neglect the influence of VM submissions and terminations, or require the workload to be specified with little abstraction. In this paper, we present a unified approach for modeling IaaS workloads. Our workload model describes the IaaS workload as a sequence of time-triggered, eventdriven external influences. We implement our model as an extension to Palladio and SimuLizar. Finally, we illustrate how historical real-world measurements are leveraged to evaluate resource management policies.

Even though the cloud era has begun almost one decade ago, many problems of the first hour are still around. Vendor lock-in and poor tool support hinder users from taking full advantage of main cloud features: dynamic and scale. This has given rise to tools that target the seamless management and orchestration of cloud applications. All these tools promise similar capabilities and are barely distinguishable what makes it hard to select the right tool. In this paper, we objectively investigate required and desired features of such tools and give a definition of them. We then select three open-source tools (Brooklyn, Cloudify, Stratos) and compare them according to the features they support using our experience gained from deploying and operating a standard three-tier application. This exercise leads to a fine-grained feature list that enables the comparison of such tools based on objective criteria as well as a rating of three popular cloud orchestration tools. In addition, it leads to the insight that the tools are on the right track, but that further development and particularly research is necessary to satisfy all demands.

Cloud computing is becoming a widely adopted paradigm for the provision of different types of ICT services. Building on a set of combined technologies, it enables service provision following on-demand usage patterns. Different vendors aim at addressing the needs of private and public large and small organizations. To this end, several solutions have been developed tackling varying service, platform and infrastructure needs. In this paper, we present the outcomes of our experimentation with respect to two open-source infrastructure-level solutions, namely OpenStack and Synnefo which have attracted the attention both of the research community and of the industrial one. The goal of our qualitative and quantitative survey and experimentation is to identify shortcomings and areas for improvement in the widely adopted OpenStack project.