Search

Virtualization technologies and server consolidation are the main drivers of high resource utilization and energy efficiency in modern Data Centers. However, some combinations of virtual machines into the same server may lead to severe performance degradation. This performance degradation is known as virtual machine interference. In a typical Data Center, different measures of virtual machine interference can be employed, depending on applications importance. Supporting a higher virtual machine interference may result in a higher consolidation, while strict low interference requirements may demand more resources. This paper presents an algorithm for server consolidation that uses an adjustable virtual machine interference threshold to map virtual machines into physical servers, allowing users to get a better trade off between amount of resources and performance according to their needs. Simulation results show that the solution succeeds in maintaining the interference levels below a defined threshold while also providing efficient server consolidation.

Vehicular Ad hoc Networks (VANETs) have been providing promising solutions for both the safety and non-safety requirements of 21st century drivers. Since, the number of vehicles are increasing, the current host centric communication approach faces various challenges. On the other hand, the Information Centric Networking (ICN) concept in vehicular communications aims to achieve content retrieval in the dynamic and error prone environment. Recently, the concept Vehicular ICN (VICN), has been explored extensively by researchers, leading to a number of open challenges such as, content naming, name resolution, Interest forwarding strategies, content store, forwarding information base and pending interest table management, security and trust issues, etc. In this paper, we therefore discuss the recent advancements in different VICN approaches with more focus on the Interest forwarding strategies which have been applied during the content retrieval process.

This paper presents an extension to the Arduino framework that introduces multitasking support and allows running multiple concurrent tasks in addition to the single execution cycle provided by the standard Arduino framework. The extension has been implemented through the ERIKA Enterprise open-source real-time kernel, while maintaining the simplicity of the programming paradigm typical of the Arduino framework. Furthermore, a support for resource sharing has also been integrated in the external Arduino libraries to guarantee mutual exclusion in such a multi-task environment.

This paper presents a new ensemble method for learning from non-stationary data streams. In these situations, massive data are constantly generated at high speed and their target function can change over time. The proposed method, named Fast Adaptive Stacking of Ensembles (FASE), uses a meta-classifier to combine the predictions from the base classifiers in the ensemble. FASE maintains a set of adaptive learners, in order to deal with concept drifting data. The new algorithm is able to process the input data in constant time and space computational complexity. It only receives as parameters the confidence level for the change detection mechanism and the number of base classifiers. These characteristics make FASE very suitable for learning from non-stationary data streams. We empirically compare the new algorithm with various state-of-the-art ensemble methods for learning in non-stationary data streams. We use a Naive Bayes classifier and a Perceptron to evaluate the performance of the algorithms over real-world datasets. The experiment results show that FASE presents higher predictive accuracy in the investigated tasks, being also able to bound its computational cost.

This paper presents a formal semantics of multicore architectures with private cache, shared memory, and instantaneous inter-core communications. The purpose of the semantics is to provide an operational understanding of how low-level read and write operations interact with caches and main memory. The semantics is based on an abstract model of cache coherence and allows formal reasoning over parallel programs that execute on any given number of cores. We prove correctness properties expressed as invariants for the preservation of program order, data-race free execution of low-level operations, and no access to stale data.

In this paper, we have defined a static approach named DUM (Detecting Unreachable Methods) that works on Java byte-code and detects unreachable methods by traversing a graph-based representation of the software to be analyzed. To assess the validity of our approach, we have implemented it in a prototype software system. Both our approach and prototype have been validated on four open-source software. Results have shown the correctness, the completeness, and the accuracy of the methods that our solution detected as unreachable. We have also compared our solution with: JTombstone and Google CodePro AnalytiX. This comparison suggested that DUM outperforms baselines.

Event-driven software systems are hard to thoroughly test because they possess a huge number of possible event sequences to test and also require significant manual effort and/or special hardware support during test case execution. In this paper we propose a novel testing approach, we call model-based replay testing that combines the advantages of model-based testing and an operating system (OS) level replay mechanism to overcome such difficulties of event-driven software testing. Specifically, we use a finite state machine (FSM) model and an n-switch set cover to automatically generate abstract test cases. We then introduce the use of OS-level replay mechanism in transforming abstract test cases into concrete test cases and executing them in an automated manner. To demonstrate the effectiveness of the proposed approach, we also describe a case study on GUI software testing using a prototype implementation of an OS-level replay mechanism in Android software platform.

Virtualization is entering the world of real-time embedded systems. Industrial automation systems, in particular, can benefit from what virtualization has to offer: flexible consolidation of applications on different hardware types and scales, extending the life-time of legacy code or decoupling software and hardware lifecycles. However, such systems require a light-weight virtualization technology in order to be able to maintain real-time behavior while dealing with real-time data. This paper sets out to investigate the applicability of container-based OS-level virtualization technology to industrial automation systems. To this end, we provide insights into the capabilities of containers to achieve flexible consolidation and easy migration of industrial automation applications as well as into the container technology readiness with respect to the fundamental requirement of industrial automation systems, namely performing timely control actions based on real-time data. Moreover, we provide an empirical study of the performance overhead introduced by containers based on micro-benchmarks that capture the characteristics of targeted industrial automation applications.

VANETs are self-organized networks in which their constituent nodes are vehicles. In general, applications developed for VANETs make use of broadcast information. However, there are many issues to be considered. One of the central problems is the broadcast storm. In this paper a temporal probabilistic protocol based on Game Theory and the Statistic ExponentiallyWeighted Moving Average (EWMA) is proposed, called E-ProbT, used to mitigate the broadcast storm problem. The Game Theory will be used to model rewards and penalties for neighbors, according to their packet forwarding decisions. The Statistic Exponentially Weighted Moving Average will be used to determinate values of threshold that indicate the best forwarders. The E-ProbT performance was measured and compared to the performance of the protocols: blind ooding, Wheighted p-Persistence, AutoCast and Irresponsible Forwarding. The results show that, according to the parameters of the E-ProbT, the delivery and redundancy rate are adjusted to desired levels.

Data streams generated in real-time can be strongly temporally dependent. In this case, standard techniques where we suppose that class labels are not correlated may produce sub-optimal performance because the assumption is incorrect. To deal with this problem, we present in this paper a new algorithm to classify temporally correlated data based on deferral learning. This approach is suitable for learning over time-varying streams. We show how simple classifiers such as Naive Bayes can boost their performance using this new meta-learning methodology. We give an empirical validation of our new algorithm over several real and artificial datasets.

Heterogeneous multi-core systems are becoming more and more common today. To be used to their full potential, the operating system has to be adapted to the new system environment. This is especially true for the scheduler as it is crucial to the overall system performance. In this paper, we present a scheduling approach for heterogeneous systems with two different kinds of cores. One that is very power efficient, but shows only a limited computing power, and the other one that has a very high performance and is very power consuming at the same time. We consider such heterogeneity for a centralized scheduler architecture. In our approach, we introduce a new load metric in order to classify tasks whether or not they are suited to be executed on a high-performance core. Based on this metric, we present a task state model for scheduling tasks according to their performance classification. We implemented the scheduling approach by extending the Brain Fuck Scheduler (BFS) and evaluated it on an eight core heterogeneous architecture with four low performance and four high-performance cores. The evaluation covers system responsiveness and high load behaviour compared to the vanilla BFS and the decentralized Completely Fair Scheduler (CFS). Even though our approach takes the heterogeneity into account, the results show that it scales better than the vanilla BFS while nearly maintaining its superior responsiveness.

Operational Transformation (OT) is a coordination approach used for supporting optimistic replication in distributed collaborative applications, where many users cooperatively manipulate replicated shared objects such as textual or multimedia documents. It allows users to concurrently update the shared data and exchange their updates in any order since the consistency of all replicas is ensured in all cases. In this context, it is desirable to undo some erroneous modifications for recovering the last correct state. This task is more tricky when out-of-order executions are considered. To avoid untimely and useless cancellations, a selective undo scheme based on OT approach is used to cancel the effect of any logged operation. Nevertheless, combining OT and undo approaches is a challenging problem since the process of doing and undoing should be reversible while preserving the data consistency. In this paper, we propose a theoretical study for this combination by formulating the undo problem as a Constraint Satisfaction Problem (CSP). For this purpose, we define a set of constraints to be fulfilled by the transformation function so that undoability is correctly managed. These constraints allow us to reduce the search space and output a reasonable number of transformation functions. The study of this output shows that commutativity somehow impacts on undoability. As the main result, to correctly manipulate undo for a shared object, we prove that the set of operations of size n • 4 must be totally commutative, while only a subset of commutative operations is necessary otherwise. As such, this work opens new perspectives for the design of safe OT-based collaborative applications using CSP methods.

Linearizability is a widely recognised correctness criterion for concurrent objects. A number of proof methods for verifying linearizability exist. In this paper, we simplify one such method with a systematic approach for invariant generation. Although this existing refinement-based method is itself systematic and fully tool-supported, it requires the verifier to provide a specific invariant over the implementation. While a chosen invariant may suffice for some proof obligations of the method, it may not for others resulting in a new, stronger invariant to be chosen and the previously completed proof steps to be redone. Our approach avoids such wasted proof effort by generating an invariant which is guaranteed to be sufficient for all proof obligations.

Recently, the notion of RAID technique has been adopted in SSD-based storage systems to improve the system reliability. However, directly applying RAID technique over SSD's would incur performance degradation due to the out-place-update nature of flash memory. Although using NVRAM as the partial parity cache could alleviate the overhead of frequent parity updates, the contribution of the partial parity cache is confined by its limited size. Some researchers proposed to use an dedicated HDD to store parity data for SSD-based RAID. However, the performance of SSD-based RAID is deteriorated by the inherent overhead of HDD, i.e., seek time and rotation latency. In this paper, we propose a parity management scheme that adopts NVRAM as partial parity cache and uses HDD as parity buffer. With proper management, the performance degradation due to limited cache size can be avoided, while the inherent overhead of HDD can be hidden. The experiment results showed that the proposed PMS-RAID could achieve up to 51.7% improvement in write throughput and up to 47.1% improvement in total block erasures, compared with RAID-5S.

Dynamic component-based platforms allow software to evolve at runtime, that is, components that can be located, loaded, and executed during runtime. Such dynamic update mechanism provides flexibility but introduces new challenges. This is especially true when dealing with third-party components, which make hard to predict the impacts (e.g., component incompatibilities at runtime, errors leading to application crashes) when integrating such thirdparty code into an application. Component quality is something hard to be evaluated and even harder when components are combined together. Third-party components whose origin or quality attributes are unknown may be considered as untrustworthy since they may potentially introduce faults to applications, although unintentionally. This paper describes the dynamic policy (i.e. changeable at runtime) behind our solution for temporarily isolating components in a sandbox, avoiding the trusted components to be disturbed in case the third-party code behaves inappropriately. By providing such mechanism we help introducing dependability attributes (namely maintainability, reliability and availability) in the component platform's architecture. In case the component presents no harm to the system, our approach provides the ability to promote a component outside the sandbox.

Affective computing in human-computer interaction research enables computers to understand human affects or emotions to provide better service. In this paper, we investigate the detection of human attention useful in intelligent e-learning applications. Our principle is to use only ubiquitous hardware available in most computer systems, namely, webcam and mouse. Information from multiple modalities is fused together for effective human attention detection. We invite human subjects to carry out experiments in reading articles being subjected to different kinds of distraction to induce different attention levels. Machine-learning techniques are applied to identify useful features to recognize human attention level. Our results indicate improved performance with multimodal inputs, suggesting an interesting affective computing direction.

This paper presents FIDE, an Integrated Development Environment (IDE) for building applications using Functional Mock-up Units (FMUs) that implement the standardized Functional Mock-up Interface (FMI). FIDE is based on the actororiented Ptolemy II framework and leverages its graphical user interface, simulation engine, and code generation feature to let a user arrange a collection of FMUs and compile them into a portable and embeddable executable that efficiently co-simulates the ensemble. The FMUs are orchestrated by a well-vetted implementation of a master algorithm (MA) that deterministically combines discrete and continuous-time dynamics. The ability to handle these interactions correctly hinges on the implementation of extensions to the FMI 2.0 standard. We explain the extensions, outline the architecture of FIDE, and show its use on a particularly challenging example that cannot be handled without the proposed extensions to FMI 2.0 for co-simulation.

Most of the supervised classification algorithms are proposed to classify newly seen instances based on their learned label space. However, in the case of data streams, concept-evolution is inevitable. In this paper we propose a support vector based approach for classification beyond the learned label space in data streams with regard to other challenges in data streams like concept-drift and infinite-length. We maintain the boundaries of observed classes through the stream by utilizing a support vector based method (SVDD). Newly arrived instances located outside these boundaries will be analyzed by constructing neighborhood graph to detect the emergence of a class beyond the learned label space (novel class). Our method is more accurate to model intricate-shape class boundaries than existing method since it utilizes support vector data description method. Dynamically maintaining boundaries by shrinking, enlarging and merging spheres in the kernel space, helps our method to adapt both dramatic and gradual changes of underlying distribution of data, and also be more memory efficient than the existing methods. Conducted experiments on both real and synthetic benchmark data sets show the superiority of the proposed method over the state-of-the-art methods in this area.

Since building software processes from scratch involves considerable effort, process tailoring should be practiced to adapt existing processes deriving new alternate ones that fit specific needs. Although several approaches have already been proposed to specify software process tailoring, some challenges remain open. Techniques and mechanisms are still required to specify and track specific variations on the software process control-flow. Thus, this paper presents a new representation structure for software process tailoring that is based on high-level operations defined as BPMN extension concepts and associated support mechanisms. Such structure intents to contribute allowing the specification, record, and traceability of detailed control-flow variations required in the context of software process tailoring as well as facilitating the specification of such variations and improving their understanding.

User profiling is an important aspect of recommender systems. It models users' preferences and is used to assess an item's relevance to a particular user. In this paper we propose a profiling approach which describes and enriches the users' preferences using multiple types of interactions. We show in our experiments that the enriched version of users' profiles is able to provide better recommendations.

User trajectories can be used to extrapolate personal information such as interests and movement patterns. Extrapolating this information is especially important in the context of opportunistic networks, which take advantage of human mobility and their interactions to deliver messages to relevant users. In this paper, we propose a Geo-casting routing protocol called LoSeRO for opportunistic networks, which uses knowledge of the locations most frequently visited by a user to route messages. LoSeRO forwards messages---in a multicasting way---to all users who have a mobility profile that intersects the packet's destination zone. In particular, users' mobility profile is based on pre-defined zones, and LoSeRO generates a mobility vector, called MobyZone, populated by the most n-frequented zones. Thus, if the destination zone of the packet belongs to the MobyZone of a user, then the user is chosen as valid candidate for receiving the packet. Efficiency of our protocol has been evaluated using different metrics, such as coverage, precision and success rate, and compared to that of state-of-the-art geo-casting protocols. Simulations show that LoSeRO reaches the best compromise among the mentioned evaluation metrics.

Distributed systems consist of a set of independent processors interconnected by a communication network that supports resource sharing. A deadlock occurs in a distributed system when a group of processes waits indefinitely for resources from each other. Distributed systems are usually represented by wait-for graphs, where the behavior of a process is determined by a deadlock model. In this paper, we revisit deadlock model concepts, and present a new deadlock model as a simpler alternative to the And/Or model. Using also computational complexity and circuit complexity aspects, we provide a novel analysis of the hierarchy of classical deadlock models, where we identify how expressive each model is from the point of view of polynomial computations. Finally we present a generic graph structure to characterize deadlock situations.

Cloud computing has revolutionized the way IT is provisioned nowadays since it exposes computing capabilities as rental resources to consumers. The emergence of cloud computing services hasn't though prevented outages in these environments even among high profile ranked cloud providers. Tremendous efforts concentrated on fault management measures have been applied for these environments. But they have been focused mainly on the IaaS service model and have been operated on the cloud provider side alone. In this context, this paper proposes an architecture for cloud brokering that implements dependability properties in an end to end way involving different cloud actors and all over the cloud service models SaaS, PaaS and IaaS.

Modern computer systems employ multiple cores and various layers of memory modules such as local memory and remote memory. Cores and memory modules are connected through diverse micro-architecture resources including LLC (Last Level Cache), IMC (Integrated Memory Controller), Interconnect, and GQ (Global Queue). Since these micro-architecture resources are shared among cores while accessing data, the interference of these resources gives a significant influence on the system performance. To explore this influence, in this paper, we observe how the execution time changes when we execute SPEC CPU2006 and PARSEC benchmarks with different data and threads placement on our experimental system. We also measure the usage behavior of the micro-architecture resources using PMU (Performance Monitoring Unit) and analyze the correlation between the execution time and PMU parameters. Our finding is that the two PMU parameters, called IMC_reads and writes and QHL_requests are good indicators whose differences are well matched with the performance differences. It implies that changing these indicators in a positive way by controlling data and/or thread placement can boost the system performance.

When testing from a deterministic Finite State Machine specification M of an implementation N, state identification plays an essential role. In the case that M does not have a single input sequence to identify its states, state identification has to be performed by using a set of input sequences, elements of which need to be applied at the same state of N. Locating sequences have been proposed in the literature in order to guarantee that N is brought back to the same state over and over, in order to apply these input sequences. Although the lengths of locating sequences are exponential in general, there are some methods proposed in the literature to reduce their length. In this paper, we present an improved method for reducing the length of locating sequences. We give a formal description of our method and explain how the existing methods relate to our method.

Along with the development of mobile devices, it became convenient to produce and consume media contents and accordingly, various media applications have been developed. Proliferation of media contents and the desire to share them with others necessitate novel mobile software technologies that enable convenient and efficient media sharing. While current mobile platforms provide lots of functionalities for media production and consumption, they do not provide any functionality for media sharing among remote users. In this paper, we design and implement a remote music play service to Android platform. Using this platform service, one can share his/her music contents with remote users without using any contents sharing application. And since local music contents are transferred to other device in a streaming, not a file transfer, manner, it is free from copyright violation problem. Experimental results show that the proposed music sharing service shows not only lower latency but also lower overall power consumption than an application level file transferbased music sharing. Using our modified Android platform, users can share their music contents with others without installing any media sharing application.

Smart City applications often use event processing techniques to detect coarse-grained events and situations from fine-grained events of the physical and social world. They operate in dynamic environments in which the properties of underlying resources and streams need to be constantly updated according to changes and events in the real world (e.g. sensor readings, network availability, weather conditions, and temperature). In most of the existing solutions matchmaking between the requirements expressed by event consumers and available event providers is carried out at design-time. This approach is often far from optimal and its deficiencies become even more obvious in smart city scenarios due to their inherently dynamic stream properties. In this paper we discuss a solution for quality-aware adaptive complex event processing using a service-oriented approach. We detail the automatic adaption strategies and evaluate them in a smart city scenario with both real and synthesised datasets.

There is growing evidence that spear phishing campaigns are increasingly pervasive, sophisticated, and remain the starting points of more advanced attacks. Current campaign identification and attribution process heavily relies on manual efforts and is inefficient in gathering intelligence in a timely manner. It is ideal that we can automatically attribute spear phishing emails to known campaigns and achieve early detection of new campaigns using limited labelled emails as the seeds. In this paper, we introduce four categories of email profiling features that capture various characteristics of spear phishing emails. Building on these features, we implement and evaluate an affinity graph based semi-supervised learning model for campaign attribution and detection. We demonstrate that our system, using only 25 labelled emails, achieves 0.9 F1 score with a 0.01 false positive rate in known campaign attribution, and is able to detect previously unknown spear phishing campaigns, achieving 100% 'darkmoon', over 97% of 'samkams' and 91% of 'bisrala' campaign detection using 246 labelled emails in our experiments.

With the increase dependence on software, there is a pressing need for engineering long-lived stable software. As architectures have a profound effect on the life-span of the software and the provisioned quality of the service, a stable architecture is a significant asset of the software. Yet, there is lack of consensus on the concept of stability as an architectural quality attribute. This paper proposes a taxonomy for defining, characterising and analysing architectural stability. The aim is to provide a better understanding and characterisation of this strategic quality attribute within the domain of software architecture, and to explicate a set of general concepts across a wide range of architectures. Such framework would significantly ease understanding the concept in the shed of modern and complex software architectures, and, therefore, allow more systematic guidance in designing and operating architectures. We apply the taxonomy in analysing the behavioural stability of self-aware cloud architectures.

The purpose of this paper is to study the scheduling problem of mixed task sets on multiprocessor systems. We propose a scheduling framework to improve the average response time of aperiodic jobs in mixed tasks. Our proposed scheduling framework consists of two phases: In the first phase, we considers how to assign each periodic task to a proper processor by different task allocation heuristics. Then, on a processor the unused capacity left by periodic tasks is set as the size of the Total Bandwidth (TB) Server. In the second phase, at run time when an aperiodic job arrives in the system, it will be assigned to a proper TB Server with the minimal virtual deadline and scheduled with periodic jobs on the same processor under Earliest Deadline First (EDF). A series of experiments were conducted to evaluate the proposed framework.

In this paper, we propose new types of cascading attacks against smart grid that use control command disaggregation and core smart grid services. Although there have been tremendous research efforts in injection attacks against the smart grid, to our knowledge most studies focus on false meter data injection, and false command and false feedback injection attacks have been scarcely investigated. In addition, control command disaggregation has not been addressed from a security point of view, in spite of the fact that it is becoming one of core concepts in the smart grid and hence analysing its security implications is crucial to the smart grid security. Our cascading attacks use false control command, false feedback or false meter data injection, and cascade the effects of such injections throughout the smart grid subsystems and components. Our analysis and evaluation results show that the proposed attacks can cause serious service disruptions in the smart grid. The evaluation has been performed on a widely used smart grid simulation platform.

This paper presents an analytical model for anycast asynchronous preamble-based medium access control (MAC) protocols. Asynchronous MAC protocols are widely used in Wireless Sensor Networks (WSN). An important drawback of these protocols is the sleep-delay phenomenon, which may impose high communication latency. In order to tackle this shortcoming, the concept of anycast, i.e., to explore opportunistic path redundancy is employed, instead of selecting a unique next hop. Existing approaches employ arbitrary rules to select the possible next relays. The mathematical model introduced in this work enables a better selection of next hop candidates based on packet size and network density, in order to minimize the sleep-delay. Along with the model, experimental results are presented validating the proposal.

In this paper, we propose an adaptive, local hole-bypassing algorithm for geographic routing in Wireless Sensor Networks (WSNs). Void areas exist in WSNs mainly because of low-density sensor-node distribution or the drain of a sensor node's energy supply. In geographic forwarding and routing protocol, when a receiving node of a packet is nearest to the packet's destination, the node declares itself a hole node and drops the packet. We therefore propose adding a mechanism in the forward-routing process, wherein the hole node forwards the packet backward, in the hopes of finding a path to the destination node. The number of times a packet can be backward-forwarded is limited to a maximum value. Herein, an adaptive mechanism is proposed to set this maximum value. We also show, through simulation, that our algorithm outperforms the baseline geographic-forwarding algorithm, as well as some fixed, maximum values over a sensor network's lifetime in terms of its packet-delivery ratio.