Many Big Data applications in science and industry have arisen, that require large amounts of streamed or event data to be analyzed with low latency. This paper presents a reactive strategy to enforce latency guarantees in data flows running on scalable Stream Processing Engines (SPEs), while minimizing resource consumption. We introduce a model for estimating the latency of a data flow, when the degrees of parallelism of the tasks within are changed. We describe how to continuously measure the necessary performance metrics for the model, and how it can be used to enforce latency guarantees, by determining appropriate scaling actions at runtime. Therefore, it leverages the elasticity inherent to common cloud technology and cluster resource management systems. We have implemented our strategy as part of the Nephele SPE. To showcase the effectiveness of our approach, we provide an experimental evaluation on a large commodity cluster, using both a synthetic workload as well as an application performing real-time sentiment analysis on real-world social media data.

Nature has provided an elegant solution for the routing problem millions of years ago, when ant colonies started to use swarm intelligence to discover food and route it reliably to their formicaries. The approach utilized by ants has several advantages that are also useful in computer networks: complete distribution, load balancing, finding shortest paths with a high probability. Several routing protocols designed for the area of computer networks have made use of this approach, called ant colony optimization. This paper provides first a broad overview of ant colony optimization-based routing protocols, while focusing on four selected approaches in later sections, describing their operation and discussing their properties in detail.

This paper presents a simulation tool integration framework which aims to integrate tools from various research fields in order to provide a seamless design flow for distributed systems with Emergent Self-Organization (ESO). ESO is a decentralized dynamic process with many favorable properties which makes it attractive for large scale distributed systems. However, ESO has a "non-linear" and "random" nature requiring expertise and tools from various research fields. Instead of extending every simulator and modeling framework to contain some of the required tools we propose a tool integration framework consisting of a repository of ESO design tools together with an integration infrastructure. The tools can be immediately used, thus significantly reducing development time and helping to fully explore the potential of ESO. Moreover, these tools can be created by experts on various aspects of ESO and seamlessly used in any relevant application domain. We demonstrate the usefulness of the proposed framework by showing how it can be used to improve an existing self-organizing data collection algorithm for Wireless Sensor Networks.

Nowadays, a range of applications and systems are emerging that consist of hundreds or thousands of computing devices, which are available at low cost but severely limited in their resources. A proper orchestration of these elements can yield a powerful system, which performs complex tasks, while staying simple and making efficient use of its resources. Designing such systems is a challenging task. The classical approach is to develop them in an ad hoc manner, which does not fully leverage the potential of these systems and is far from what can be achieved. There are methodologies which try to overcome this challenge but they do not address essential issues. Moreover, most of them do not provide a design environment to support the design process. In this paper, we propose a new methodology and design environment for systematic construction of emergent behavior in large-scale distributed embedded systems which address issues at both micro- and macroscopic layer. Our approach and simulator were successfully used to design emergent clustering and data collection in large scale wireless sensor networks.

Discrete-event wireless network simulators are important tools, which enable an efficient development and evaluation of communication protocols. They can be used already at early project stages and provide useful results, when e.g. a real-world deployment or mathematical analysis is not applicable. One of the major challenges in the field of discreteevent wireless network simulators is the relatively high amount of time needed to run simulations. As scenarios and protocols become more complex, the time needed to run simulations can amount to several hours for a single run. At the same time, tens of runs are needed to produce statistically significant results only for one parameter combination, which leads to total simulation run times of several days or even a few weeks, since usually several parameter combinations are needed. Therefore, many developers distribute their simulations manually, which however often yields poor solutions, since only a small portion of their time is available for this task. To cope with this problem, in this paper, we present a distributed environment which extends the ShoX network simulator. It distributes a set of runs automatically and manages the overall simulation execution. The user can determine the nodes on which the simulations are executed, the parameter and seed ranges, among other settings. In order to improve the ease of use, our approach offers a GUI and collects the simulation results at a specified target.

Massively Distributed Embedded Systems (MDES) such as Wireless Sensor Networks (WSN) are gaining increasing attention, since they enable a broad range of novel applications starting from monitoring oceans to exploring distant planets. WSNs consist of hundreds of nodes that have typically very limited recourses (computational, memory, energy, etc.) and are deployed in a dynamic environment, where they have to continuously adapt to new conditions. Due to the small-size requirement of the nodes, they are highly resource-constrained. Because of that, the amount of functionality that may be present in each node is limited. Therefore, cooperation between nodes is needed in order to accomplish complex tasks. These facts turn the design of applications for WSNs into a challenge. A promising approach how to deal with it is to use the emergent self-organization metaphor. In this paper a new process algebra (PA) called "Locality-aware extension of π-Calculus" is presented. The algebra is one of several techniques included in a new design methodology for the design of self-organizing behavior in MDES. The method is based on π-Calculus and allows a highlevel description of interactions among processes. As the most important characteristic of self-organization is the restriction of interactions to neighboring elements (localized interactions), we extend the π-Calculus with locality awareness, a necessary abstraction to allow the modeling of self-organization in MDES. To get full locality awareness in π-Calculus, we extended it with concepts for modeling spatiality, probability and time. Moreover, new types of channels are included to cover various types of communication such as distribution, broadcast and aggregation. In order to validate this new PA, we successfully model a self-organizing clustering algorithm for WSNs.

Given the resource constraints and the relatively high cost of communication in wireless sensor networks, data collection presents one of the central challenges in this network class. In this paper, we propose an efficient self-organizing data collection protocol, which addresses this challenge. Exploiting the knowledge of the underlying topology, the proposed approach uses a novel link acceptor, responsible for choosing the links to be added to the data routing structure, as well as, a next-hop selector taking into account energy information and the distance towards the sink. Using this information, the introduced next-hop selector allows a straightforward adjustment of the trade-off between preferring shorter routes or a higher degree of load balancing. Further, the amount of oscillations is minimized by using polarization functions and probabilistic next-hop selections.

In this paper, we present a novel multi-view visualization tool for trace files generated by network simulations, currently supporting the ns-2 and ShoX trace formats. As the first tool in this area, it shows simultaneously three synchronized views (or perspectives) of the current network state in one window: (a) An enriched text view showing events line-by-line, marked up using colors and symbols in order to improve readability, (b) the 2-D bird's eye view known from other simulators, (c) a newly proposed view depicting the development of network state as a graph focusing on the visualization of communication patterns on intra- and internode level. Using this synchronized multi-view, as well as, our novel graph view, we aim at increasing the speed of the analysis and evaluation processes during the development of communication protocols.

Although network simulators present an effective means for the evaluation of communication protocols, there are still several unresolved challenges. To address them, in this paper, we provide an overview of different tools devised by us that aim to improve the efficiency of the protocol development and evaluation process in the ShoX network simulator. The presented tools include a (a) novel multi-view visualization that combines an enriched text view with a 2-D bird’s eye and a graph view, enabling the analysis of communication patterns; (b) a protocol-state visualization allowing a flexible mapping between state and representation; (c) a simulation execution environment offering a user-friendly GUI enabling the distribution of large amounts of simulation runs; and (d) a statistics generation component, which provides a straightforward means to create statistics charts. The discussed tools were implemented, evaluated, and are available as open source within the ShoX network simulator project.

Distributed RSoCs are the next step towards a new generation of embedded systems. Applications running on heterogeneous distributed RSoCs require an OS which dynamically adapts to their variable demands. In this paper, we present a novel decentralized OS service design, which enables OS adaptiveness, resource sharing, and reconfigurability on distributed RSoCs. The challenges faced by this design are classified and discussed. To cope with them, biologically inspired algorithms, e.g. for service discovery, are adopted, which use only local information provided by an RSoC and its direct neighbors.

Wireless sensor networks (WSN) are rapidly gaining attention since enabling a broad spectrum of novel applications ranging from vineyard monitoring to space exploration. WSNs may consist of up to thousands of nodes, which typically have limited recourses (computational, memory, energy, etc.) and are deployed in a dynamic environment where they have to continuously adapt to new conditions. Given the resource constrains of typical WSN nodes, the amount of functionality that can be realized at each node is also highly limited. Therefore, cooperation between nodes is needed in order to accomplish more complex tasks. These facts turn design of applications for WSNs into a challenging endeavor. A promising approach how to cope with this is using the emergent self-organization metaphor. Unfortunately most of the current designs using this metaphor are developed employing the ad hoc and trial-and-error method, which is inherently unsystematic, as well as, inefficient. In this paper we present an approach how to systematically design emergent self-organization for massively distributed embedded systems like WSN.

Connected k-hop dominating sets (CkDS) have various applications in wireless sensor networks. They are for example used to alleviate the broadcast storm problem or to achieve an adjustable amount of area coverage while maintaining connectivity. In this paper, we propose a novel protocol for the construction of CkDS in wireless sensor networks drawing inspiration from the flight behavior of ovipositing Pieris rapae, which has several properties that are beneficial for our artificial system. Our protocol is self-organizing since a global-level pattern, the CkDS, emerges solely from numerous lower-level interactions specified by rules executed using only local information, without reference to the global pattern. It consists of two intertwined behavior blocks, which are both essentially based on random walks: the first is responsible for the construction of a k-hop dominating set, while the second connects the existing fragments of dominating nodes to a connected k-hop dominating set. We conducted extensive simulations to evaluate the efficiency as well as other properties, such as scalability, of our protocol and to compare it to a recently proposed state-of-the-art CkDS construction protocol.

For the evaluation of wireless network protocols, simulation has established itself as a good compromise between cost and complexity, on the one hand, and accuracy of results, on the other hand. Most established network simulators, however, were originally developed for wired networks and only later extended to the wireless domain. Some simulators like Opnet are not even restricted to communication networks at all, but can be used to simulate all kinds of networked systems. In this paper, we present ShoX, a network simulator explicitly designed for the simulation of wireless networks. Together with its strong GUI support, this allows a very efficient development and evaluation of network protocols without any considerable learning curve. All relevant concepts known from the domain of wireless networks like nodes, links, OSI layers, packets, mobility, signal propagation and traffic models are directly available. Defining new layers, packets or models is as easy as subclassing the existing abstract classes which ensures at compile time that all required methods are provided by the user. ShoX is completely based on Java and XML, thus superseding to learn unfamiliar languages like Tcl or NED. It features an integrated GUI for configuration, visualization and statistics.

Designing protocols for wireless networks is a challenging task. Combined with the fact that such networks are often deployed for critical missions like forest fire detection in the WSN scenario or have to function properly and efficiently for an extended period of time, it is desirable to thoroughly test, analyze and evaluate newly developed communication protocols before deployment. In order to do this, simulations are a good compromise between cost/complexity and accuracy of the results. Since there are many simulators for wireless networks, it is often difficult to decide which one to choose. Therefore, we present a case study in which four common wireless network simulators were used to evaluate a well-known topology control protocol (SPAN). Within the case study, we describe the strengths and weaknesses of the examined network simulators: First, we evaluate the usability of the simulators in terms of different parts of the protocol developer's work process. Moreover, we also focus on the simulator's support for reusability and maintainability of simulation models by measuring particular model properties. For this purpose, we have proposed a model of quality for network simulators. The model of quality defines which properties of models to measure and how to interpret them. As opposed to other simulator comparisons, we do not focus on the correlation of the individual simulation results. Through this paper, we aim at providing a basis for finding an adequate simulator for a particular task.

We propose to make use of a completely distributed way of implementing OS services for wireless sensor networks (WSN). I.e. instead of having an instance of the OS on each node of a WSN the services of the OS are distributed over the WSN. Of course this approach implies specific challenges. Two of them are discussed in the paper: Migration of services to nodes such that the overall communication costs are minimized and forming clusters with the tendency to concentrate service requests inside the clusters and at the same time minimizing intra-cluster communication. For both problems biologically inspired solutions are discussed. Service migration is mapped on an Ant Colony Optimization (ACO) technique while as a clustering heuristics Division of Labor in swarms of social insects is used.

In order to evaluate the behavior and performance of protocols for wireless networks, simulations are a good compromise between cost and complexity, on the one hand, and accuracy of the results, on the other hand. Since there are many simulators for wireless networks, it is often difficult to decide which simulator to choose. To help shed light on this issue, we present a case study in which four popular wireless network simulators were used to evaluate a well-known topology control protocol (SPAN). Within the case study, we describe outstanding and desirable but missing features of the simulators, outlining their strengths and weaknesses. Further, we compare the amount of effort needed for installation, familiarization, implementation (needed lines of code and lines for configuration) and visualization. As opposed to other simulator comparisons, we do not focus on the correlation of the individual simulation results, but try to compare the simulators from feature and usability point of view. This paper can help other researchers to quickly identify which simulator is most suitable for their needs.

Given the scarcity of energy in wireless sensor networks (WSNs), in-network data processing by distributed, cooperating services is often used to reduce the amount of information that has to be routed to the base station and thereby to reduce communication and energy consumption. However, to minimize the amount of communication between services and their requesters, the locations of services in the network have to be selected carefully. Therefore, this paper proposes an efficient biologically-inspired heuristic for service assignment in WSNs. In order to reduce the amount of information exchange necessary for our heuristic, we use a concept observed in ant colonies that utilizes only local information. We model packets as ants (depositing pheromones at the visited nodes), services as food sources and requesters as formicaries. To optimize an objective function (reduction of communication distance between services and requesters), an explorer agent makes local service assignment decisions based on solely local information: the pheromones deposited by the ants. Furthermore, our paper presents the formal definition of the problem of service assignment and a thorough analysis and discussion of the results of our experiments, which show the efficiency of our approach.

Future MPSoCs have to cope with unreliable functionality of their nanometer-scale internal components, mainly due to increasing sensitivity for technology parameter variations and natural radiation. While error correction is widely known in memory design, protection for arithmetic logic units within CPUs is an issue for future research. For on-chip communication resources there are many error protection techniques available, but a tradeoff has to be found between techniques on various levels. In this paper we present our concept for Autonomic MPSoCs with capabilities for runtime detection and correction of sporadic errors, and adaptation of performance, power, and dependability on changing environmental conditions. Functional elements of the MPSoC are continuously monitored and controlled by autonomic elements. Re-distribution of tasks is supported by run-time performance analysis and an autonomic operating system. Life time dependability of the MPSoC is introduced in the design optimization process.