no official weekly meeting

Privacy and Security issues in IPv6 Networks

Privacy is a very important element in every one's everyday life and it is more complicated than before because of the use of many new techniques for storing data or sharing data among many people, at the same time, over the Internet. The purpose of this talk is to explain the existing approaches in use to maintain privacy and security in IPv6 networks and then introduces the problems with the use of these mechanisms. Finally the proposed approaches and future trends to address these issues will be introduced.

Improvements and Further Challenges in the Visualization of Varying Hierarchical Datasets with Treemaps

Space-restricted techniques for visualizing hierarchies generally achieve high scalability and readability (e.g., tree maps, bundle views, sunburst). However, the visualization layout directly depends on the hierarchy, that is, small changes to the hierarchy can cause wide-ranging changes to the layout. For this reason, it is difficult to use these techniques to compare similar variants of a hierarchy because users are confronted with layouts that do not expose the expected similarity. Voronoi treemaps appear to be promising candidates to overcome this limitation. However, existing Voronoi treemap algorithms do not provide deterministic layouts or assume a fixed hierarchy. In this talk I present an extended layout algorithm for Voronoi treemaps that provides a high degree of layout similiarity for varying hierarchies, such as software-system hierarchies. The implementation uses a deterministic initial-distribution approach that reduces the variation in node positioning even if changes in the underlying hierarchy data occur. Compared to existing layout algorithms, the algorithm achieves lower error rates with respect to node areas in the case of weighted Voronoi diagrams, which is shown in a comparative study.

Built-in Recovery Support for Explorative Programming Tasks

I will present built-in recovery support as an alternative to deal with recovery. Programmers need to recover when changes suddenly reveal implications that are not only unexpected but also undesired. Because recovering from unanticipated situations often involves expensive and tedious work, literature recommends the avoidance of recovery scenarios by following a structured and disciplined approach. Best practices include to work only on one thing at a time, to perform baby steps, as well as to make proper use of versioning and testing tools. However, recovery avoidance relies on foresight and self-control, which renders the approach exhaustive and error-prone.

I have developed a tool suite called CoExist, which makes recovery fast and easy to accomplish, even if explicit commits are unavailable and tests have been ignored for a while. Moreover, when mistakes and unexpected insights are no longer associated with tedious compensation efforts, programmers are encouraged to change source code as a means to reason about it, as opposed to making changes only after evaluating and structuring them mentally. The results of a controlled lab suggest that built-in recovery support such as CoExist increases programming performance in explorative programming tasks.

Self-Service Business Intelligence

I will present some work I did before joining the HPI Research School.

Nowadays business intelligence (or data analysis) tools advertise with fancy visualizations or fast response time. But one aspect usually is neglected: How do I actually get the data that I need? The Self-Service Business Intelligence project provides a graphical query editor that enables an ordinary end-user to write queries without help of an IT specialist and without having knowledge in textual query languages.

During the presentation I want to explain basic implementation details – like data models, query generation, and query transformation – and discuss some challenges that occur while having inexperienced end-users working with your application.

Extending Process Architectures with Data Dependencies

To a growing degree the focus in Business Process Management shifts from single processes to process interactions. Business process architectures (BPA) were established as convenient way to model and analyze such interactions on an abstract level showing only message and trigger relations between the involved processes. Previously, our group presented means for formal verification of BPAs and algorithms to derive BPAs from Process Model Collections.

However, some interrelations are only given implicit through the use of data objects. In the BPMN modeling language the construct of data annotations allows to capture the effect of activities on data objects. Shared data objects are hence a means of interrelating processes. In this presentation I show how to extract hidden data dependencies between processes given process models with data annotations and object life cycles of the involved data object. The information about data dependencies is used to construct a business process architecture, thus enabling analysis with existing methods.

no official weekly meeting

Collaboration in Cyber-Physical Systems

Within the presentation, I will give an overview about my research area and discuss in detail two related work approaches about collaboration and self-adaptive systems. I show different collaboration pattern and discuss how they influence my work. Furthermore, I show some insights of my current research work by outlining problems, discuss design decisions for modeling collaboration in cyber-physical systems as first class entities and present a running example that illustrate the current state of my modeling approach.

Babelsberg/JS, a system to enable constraints in the Lively Kernel environment

The talk will consist of a short recap of my work, but I will focus on some code examples and live demos which show how I think Babelsberg in JavaScript can be used to create interactive applications. I will highlight one of the core concepts of Babelsberg -- a declarative semantics for assignments -- and present a work-in-progress more formal description of assignments.

cancelled due to illness

no official weekly meeting

no official weekly meeting

Concurrent Execution of Heterogeneous Queries on In-Memory Databases

In-memory databases management systems (IMDBMS) that leverage column-oriented storage have been proposed to run analytical queries directly on the transactional database schema, thereby reducing the execution time of even complex analytical queries to seconds. The two main reasons for this dramatic performance increase are massive intra-query parallelism on many-core CPUs and primary data storage in main memory. The benefits of these IMDBMS for enterprises are huge: analytical applications become largely independent of data staging delays, opening the way for real-time analytics. However, this very much desirable approach of processing transactional queries, classical batch-oriented analytics, as well as real-time analytics on a single database system will only be adopted in multiuser enterprise scenarios, if some form of workload management can ensure that dynamically arriving high priority queries are executed in a timely manner and that single users cannot congest the entire DBMS. Solutions for workload management have been proposed for disk-based database management systems and they typically rely on admission control and a query execution strategy that assigns each query to a single thread and multiplexes these threads on a number of processing units. These approaches are unfavorable for in-memory databases on multi-core CPUs that heavily rely on parallelism in query execution to achieve short response times for single queries. Consequently the presented work deals with query execution strategies for heterogeneous workloads in an IMDBMS that allow prioritizing certain query classes, as well as a fair resource shares among users, while executing queries efficiently and in parallel on multiple processing units. The general approach is to split incoming queries into tasks that are assigned to hardware contexts by a user-lever scheduler. Based on this concept, scheduling policies to enforce the mentioned desired system characteristics are proposed and their effectiveness is demonstrated.

Modeling Interestingness and Serendipity in Relevance Search

Popular recommendation algorithms such as Collaborative Filtering, Latent Semantic Analysis, or Locality-Sensitive Hashing often operate under the assumption that similar users like similar items and that user who like a specific item will also like similar items (i.e., items with similar features). Therefore, these algorithms determine the most similar users / items and then recommend the top-k, i.e., the k most similar users or items.

Scalable Simulation of Configurable, Sequence-dependent Fault Trees

Fault trees are a graphical modelling language for describing the failure behaviour of complex systems. While sequence-dependent behaviour makes fault tree models more realistic and is desirable especially in the context of software fault trees, obtaining failure probabilities from sequence-dependent models is computationally hard and leads to state space explosion. My master Thesis presents a more efficient, scalable simulation-based approach: Configurable, sequence-dependent fault trees are mapped to stochastic petri nets (SPN). A Monte Carlo simulation of these petri nets then yields approximations of dependability metrics.

<string>Cell and Software: Is Evolution a Software Engineer?

In the past few decades, several scientists have successfully replicated the function of biological systems into efficient and effective algorithms, which now constitute the field of Biologically Inspired Algorithms. Moreover, curiosity about the unknown biological mechanisms have led scientists to develop mathematical models that approximate biological behavior in machines.

In addition to these two aspects of biological inspiration, there are cases where scientists/engineers have independently developed approaches, later found to have already existed in cellular systems. In this talk, I will present a few such approaches that highlight striking similarities between the design and function of cells, and the design and function of software systems. To the best of my knowledge, these similarities have not been described in existing literature.

I’ll start by presenting a few selected milestones in the field of biologically inspired computing, and then introduce you to the similarities I’ve found. I’ll conclude by sharing my perspective on the observations in the form of conjecture and open questions.

Update on Muscle-IO

On this talk we follow the current trend of scaling down hardware devices and interfaces -- from desktop to laptop, smartphone to smartwatch -- which is accompanied by a reduction of the visual interfaces and brings new challenges to the field of human computer interaction.

We show an update on our current research, Muscle-IO, which is a wearable input and output device designed for intermitted micro-interactions. Muscle-IO device receives input from the user by reading the tension of the hand muscles, and produces output by writing to the user’s arm muscles. Based on this design, users enter information by performing an input gesture, such as flexing their hands inwards; and, receive output by finding their arm posed in an output gesture, such as having their palm extended outwards. Muscle-IO input and output thus share the same channel, i.e., user’s muscles. As users not only enter information eyes-free using gestures, but also perceive Muscle-IO’s response eyes free by means of proprioception, they find their hands empty and their eyes free, allowing them to focus on their primary task, such as having a conversation and keeping eye contact with their audience. When using Muscle-IO, the device allows users to switch their attention to an urgent primary task anytime and instantly by abruptly moving their arm, which also automatically dismisses