Soft sensors are mathematical models of industrial processes that are often used for monitoring and control. Data-driven techniques based on artificial intelligence are generally used for identifying these models. Data scarcity is a challenging problem that occurs when the variable to be estimated must be measured by laboratory analysis. Here, a difference-based neural network called Δ -Net is proposed to develop dynamic nonlinear soft sensors when only a few hundred labeled data are available. The Δ -Net, which exploits the case difference heuristic approach, is based on two parallel neural networks responsible for processing pairs of input samples. The Δ -Net is trained on an augmented dataset obtained by pairwise ordering the small original dataset. The outputs of the subnets are then combined to estimate the difference between the corresponding outputs of the pairs. The reconstruction of the output sample is computed as the mean of the distances from the sample to a set of anchor points, to increase the robustness of the prediction. The proposed approach has been applied to well-known industrial benchmarking datasets. The obtained results show superior performance compared to other data augmentation approaches, including bootstrap resampling, variational autoencoders and Wasserstein generative adversarial networks.

Spiking Neural Networks (SNNs) represent a novel class of models that are becoming increasingly popular. Thanks to their unique way to process discrete and asynchronous signals, these type of networks demonstrated to be computationally and energetically more efficient than traditional Deep Neural Networks (DNNs), making them suitable for devices with limited hardware capabilities. Moreover, their ability to continuously adapt to external inputs changes enables an online evolution of network weights which is ideal for real-time applications scenarios. As a downside, one of the main challenges when working with SNNs consists in the impossibility of using traditional learning rules such as backpropagation (or its variations) for the online training of these architectures. In such a context emerges the need to define online and biologically plausible training solutions that would allow the hardware implementation of these networks on neuromorphic devices. To overcome this issue, in this paper we propose a learning rule for the training of vanilla Spiking Autoencoders (Spiking-AEs) which exploits synaptic trace activities to enable an online weights adaptation. We also propose a novel modular tied Spiking-AE to address reconstruction and anomaly detection tasks. In this sense, the tied connections make the training less complex, drastically reducing the number of trainable parameters and model memory footprint. Experimental results conducted on the MNIST and MFPT datasets demonstrate the effectiveness of the proposed solution in terms of reconstruction and anomaly detection capabilities, respectively.

In the field of robotics, ensuring reliable and efficient performance is crucial, especially when robots are entrusted with critical tasks. Anomaly detection systems play a crucial role in maintaining this reliability by detecting deviations from normal behavior and taking timely interventions. Traditional model-and knowledge-based approaches, while effective in controlled environments, reach their limits in dynamic and resource-constrained settings due to their reliance on predefined models, expert knowledge and high computational requirements. While data-driven methods, especially those using deep learning, offer better adaptability, they also bring challenges in terms of energy consumption and hardware limitations. To address these issues, this paper proposes a hybrid anomaly detection system that utilizes Spiking Neural Networks (SNNs) and Convolutional Neural Networks (CNNs). The SNN provides low-power processing of input features for anomaly detection, while the CNN efficiently extracts spatial features from high-resolution images to realize an accurate, high-performance classifier that can be used to perform an online fine-tuning of the SNN. A large dataset for robot navigation is used as a testbed. The obtained results show that this hybrid approach increases the speed and accuracy of anomaly detection while significantly reducing energy consumption, making it well-suited for applications in resource-constrained robotic systems.

The tenth IEEE International Conference on Smart Computing (SmartComp 2024), sponsored by the IEEE Computer Society, will be held in-person in Osaka, Japan. It continues the tradition of the previous editions in presenting high-quality research and technology in smart and connected computing.

Anomaly detection systems are crucial for identifying deviations from normal behavior in various domains, ranging from cybersecurity to industrial system monitoring. Traditional approaches often rely on predefined rules or supervised learning algorithms, which may struggle to adapt to complex and evolving patterns. As a result, deep learning techniques have been employed to address these challenges, but they usually require significant computational resources and memory for training and inference. On the contrary, Spiking Neural Networks (SNNs), inspired by the brain neural architecture, excel in effectively capturing complex temporal patterns present in real-world data through unsupervised learning, potentially leading to more robust performance while reducing energy consumption. In this paper, we employ the homeostatic plasticity characteristic that governs the synapses behaviour in SNN to implement an anomaly detection system designed for processing raw vibration data. Initially, we generate spike trains to represent the input data. These are then fed into a Balanced Spiking Neural Network (BSNN) to detect suspicious anomalies. The proposed SNN learning algorithm dynamically adjusts synaptic weights based on the precise timing of spikes between neurons, and introduces a combination of Spike Timing Dependent Plasticity (STDP) and reverse STDP techniques. Experimental results conducted on a public available vibration dataset demonstrate the effectiveness of the proposed approach in accurately identifying anomalies while maintaining low false positive rates.

In the evolving landscape of healthcare, Machine Learning (ML) and Deep Learning (DL) have revolutionized medical practises, particularly in Smart Health. This paper explores their application in Intelligent Cyber Physical Systems (iCPSs), focusing on surface Electromyography (sEMG) signal classification for prosthetic control. We conduct a detailed quan-titative study, comparing the performances and memory efficiency of Recurrent Neural Networks (RNNs) with Long Short-Term Memory (LSTM) cells and Convolutional Neural Networks (CNNs) with Temporal Convolutional Networks (TCNs) layers. Using the most popular quantization frameworks, TensorFlow Lite (TFLite) and Quantized Keras (QKeras), we quantify the impact on the proposed models of both INT8 Quantization and Deep Quantization. The aim is to undertake a comparison among the proposed structures and prove that their Deeply Quantized version lead to minimal performance degradation related to memory footprint's substantial reduction. Our contributions include a custom TCN layer, compliant with QKeras deep quantization, a comparative analysis between LSTM and TCN and evaluation of quantization frameworks. We offer insights into model performance and efficiency, informing their potential in real-world medical contexts.

Deliver intelligence into low-cost hardware e.g., Microcontroller Units (MCUs) for the realization of low-power tailored applications nowadays is an emerging research area. However, the training of deep learning models on embedded systems is still challenging mainly due to their low amount of memory, available energy, and computing power which significantly limit the complexity of the tasks that can be executed, thus making impossible use of traditional training algorithms such as backpropagation (BP). During these years techniques such as weights compression and quantization have emerged as solutions, but they only address the inference phase. Forward-Forward (FF) is a novel training algorithm that has been recently proposed as a possible alternative to BP when the available resources are limited. This is achieved by training the layers of a neural network separately, thus reducing the required energy and memory. In this paper, we propose µ-FF, a variation of the original FF which tackles the training process with a multivariate Ridge regression approach and allows to find closed-form solution by using the Mean Squared Error (MSE) as loss function. Such an approach does not use BP and does not need to compute gradients, thus saving memory and computing resources to enable the on-device training directly on MCUs of the STM32 family. Experimental results conducted on the Fashion-MNIST dataset demonstrate the effectiveness of the proposed approach in terms of memory and accuracy.

The industrial Internet of Things technology had a very strong impact on the realization of smart frameworks for detecting anomalous behaviors that could be potentially dangerous to a system. In this regard, most of the existing solutions involve the use of artificial intelligence models running on edge devices, such as intelligent cyber physical systems typically equipped with sensing and actuating capabilities. However, the hardware restrictions of these devices make the implementation of an effective anomaly detection algorithm quite challenging. Considering an industrial scenario, where signals in the form of multivariate time-series should be analyzed to perform a diagnosis, echo state networks (ESNs) are a valid solution to bring the power of neural networks into low complexity models meeting the resource constraints. On the other hand, the use of such a technique has some limitations when applied in unsupervised contexts. In this article, we propose a novel model that combines ESNs and autoencoders (ESN-AE) for the detection of anomalies in industrial systems. Unlike the ESN-AE models presented in the literature, our approach decouples the encoding and decoding steps and allows the optimization of both the processes while performing the dimensionality reduction. Experiments demonstrate that our solution outperforms other machine learning approaches and techniques we found in the literature resulting also in the best tradeoff in terms of memory footprint and inference time.

Vehicular Social Networks (VSNs) is an emerging communication paradigm, derived by merging the concepts of Online Social Networks (OSNs) and Vehicular Ad-hoc Networks (VANETs). Due to the lack of robust authentication mechanisms, social-based vehicular applications are vulnerable to numerous attacks including the generation of sybil entities in the networks. We address this important issue in vehicular crowdsourcing campaigns where sybils are usually employed to increase their influence and worsen the functioning of the system. In particular, we propose a novel User Recruitment Policy (URP) that, after extracting the participants within the event radius of a crowdsourcing campaign, detects and filters out the sybil vehicles by using a novel sybil detection approach, called SybilDriver. This technique combines the advantages of VANETs and OSNs by means of an innovative concept of proximity graph obtained from the physical vehicular network, in conjunction with a community detection and Random Forest techniques adopted in the OSN domain. Detailed experimental evaluations demonstrate the effectiveness of our approach and also show that it outperforms existing state-of-the-art methods typically used in the OSNs.1

The Industry 4.0 paradigm has changed the way industrial systems with hundreds of sensor-actuator enabled devices, including industrial internet of things (IIoT), cooperate and communicate with the physical and human worlds. Given the intricacy, the diagnostics of such systems is extremely important. While anomaly detection is a valid approach to avoid unplanned maintenance or even complete breakdown, its effective realization in IIoT requires the design and implementation of frameworks for efficient monitoring, data collection, and analysis. Most of the existing anomaly detection techniques provide only a diagnosis of the fault without taking into account the uncertainty. Moreover, the lack of ground truth data (which is a typical problem in the industrial context), make their implementation even more challenging. This paper proposes an anomaly detection technique built on top of an industrial framework for the data collection and monitoring. Specifically, we address the lack of labeled data by designing a semi-supervised anomaly detection algorithm that exploits Bayesian Gaussian Mixtures to assess the working condition of the plant while measuring the uncertainty during the diagnosis process and we implement the proposed framework on a real-life IIoT testbed, namely a scale replica assembly plant. Experimental results demonstrate that our anomaly detection algorithm is able to detect the plant working conditions with 99.8% of accuracy, and the semi-supervised approach performs better than a supervised one.

With the advent of smart environments the requirements for diagnostics and prognostics techniques gained a lot of interest. Lately, the Industry 4.0 paradigm is starting to expand also in the smart road, referred to in this context as "maintenance 4.0". In such a context, the possibility to predict the conditions of a road network allows to deliver a preventive maintenance that can strongly reduce the costs and avoid severe consequences. However, considering the actual pavement management state of the art, it is evident the huge amount of heterogeneous data necessary to perform this challenging tasks. Leveraging the Cloud and Edge technologies, this paper proposes a web Geographical Information System (GIS) platform capable to dynamically collect and analyze several type of sensor data for the management of road pavements. On top of that, the platform is able to compute advanced indexes and metrics that are used to produce a maintenance proposal scheme. Experimental results present a preliminary case study on a real motorway and demonstrate the effectiveness of the proposed platform as a support tool during the maintenance process.

Summary The Internet of Things is underpinned by the global penetration of network‐connected smart devices continuously generating extreme amounts of raw data to be processed in a timely manner. Supported by Cloud and Fog/Edge infrastructures – on the one hand, and Big Data processing techniques – on the other, existing approaches, however, primarily adopt a vertical offloading model that is heavily dependent on the underlying network bandwidth. That is, (constrained) network communication remains the main limitation to achieve truly agile IoT data management and processing. This paper aims to bridge this gap by defining Clustered Edge Computing – a new approach to enable rapid data processing at the very edge of the IoT network by clustering edge devices into fully functional decentralized ensembles, capable of workload distribution and balancing to accomplish relatively complex computational tasks. This paper also proposes ECStream Processing that implements Clustered Edge Computing using Stream Processing techniques to enable dynamic in‐memory computation close to the data source. By spreading the workload among a cluster of collocated edge devices to process data in parallel, the proposed approach aims to improve performance, thereby supporting agile data management. The experimental results confirm that such a distributed in‐memory approach to data processing at the very edge of an IoT network can outperform currently adopted Cloud‐enabled architectures, and has the potential to address a wide range of IoT‐related data‐intensive time‐critical scenarios.

Accurate Remaining Useful Life (RUL) prediction of machines is important for condition-based maintenance, in order to improve the reliability and costs of maintenance. The rotor is one of the most important equipment parts and is one of the most common failure points. To assess the degradation life of rotating machines, this paper proposes a data-driven prognostic technique that utilizes an unsupervised trend extraction and an exponential degradation model to obtain an accurate RUL prediction of rotor failures. The main steps of the proposed prognostic technique are condition monitoring data acquisition, feature extraction using signal processing techniques, feature selection based on the monotonicity technique to quantify the merit of the best representative features for prognosis purposes. The selected features are then combined using the principal component analysis technique to get the most appropriate component health indicator. Finally, an exponential degradation model is used to estimate the RUL using this health indicator. The suitability of the proposed approach in component monitoring is demonstrated on a degradation dataset of a faulty rotor acquired from a Simulink model of an induction motor.

Diagnosis of plant health conditions is gaining significant attention in smart agriculture. Timely recognition of early symptoms of a disease can help avoid the spread of epidemics on the plantations. In this regard, most of the existing solutions use some AI techniques on smart edge devices (IoTs or intelligent Cyber Physical Systems), typically equipped with a hardware like sensors and actuators. However, the resource constraints on such devices like energy (power), memory and computation capability, make the execution of complex operations and AI algorithms (neural network models) for disease detection quite challenging. To this end, compression and quantization techniques offer viable solutions to reduce the memory footprint of neural networks while maximizing performance on the constrained devices. In this paper, we realized a real intelligent CPS on top of which we implemented an AI application, called Deep Leaf running on a microcontroller of the STM32 family, to detect coffee plant diseases with the help of a Quantized Convolutional Neural Network (Q-CNN) model. We present a quantitative analysis of Deep Leaf by comparing five different deep learning models: a 32-bit floating point model, a compressed model, and three different types of quantized models exhibiting differences in terms of accuracy, memory utilization, average inference time, and energy consumption. Experimental results show that the proposed Deep Leaf detector is able to correctly classify the plant health condition with an accuracy of 96%, thus demonstrating the feasibility of our approach on a Smart Edge platform.

Considering the proliferation of smart devices connected to the Internet, typically going under the aegis of Internet of Things (IoT), a trend has arisen to promote the Cloud paradigm as a suitable management system for such a complex environment. In this context, an effort to extend the OpenStack ecosystem to make it able to support the management of the IoT infrastructure has been made by virtue of the I/Ocloud approach, leading up to its reference implementation, the Stack4Things (S4T) middleware. S4T provides a set of suitable capabilities and features to make the (remote) IoT devices able to join an edge-based IaaS/PaaS Cloud. In the interest of enhancing the S4T middleware scalability and explore new capabilities in particular, ones related to Fog and Edge paradigms, it is becoming a must to test new features in practice at a low financial cost and particular constraints for instance, number/type of devices, network conditions, etc. For this purpose, the use of network emulation tools is a practical and suitable approach. In this paper, we present an integration between the S4T middleware and an emulation tool namely Containernet. Through the integration approach, we model network conditions (e.g., latency, bandwidth, packet loss) and devices (in forms of containers) using Containernet, and we manage the devices (i.e., containers) by means of S4T.

Aspects related to the maintenance scheduling have become a crucial problem especially in those sectors where the fault of a component can compromise the operation of the entire system, or the life of a human being. Current systems have the ability to warn only when the failure has occurred causing, in the worst case, an offline period that can cost a lot in terms of money, time, and security. Recently, new ways to address the problem have been proposed thanks to the support of machine learning techniques, with the aim to predict the Remaining Useful Life (RUL) of a system by correlating the data coming from a set of sensors attached to several components. In this paper, we present a machine learning approach by using LSTM networks in order to demonstrate that they can be considered a feasible technique to analyze the "history" of a system in order to predict the RUL. Moreover, we propose a technique for the tuning of LSTM networks hyperparameters. In order to train the models, we used a dataset provided by NASA containing a set of sensors measurements of jet engines. Finally, we show the results and make comparisons with other machine learning techniques and models we found in the literature.

Prediction-as-a-Service is a promising new paradigm that brings the advantages of Software-as-a-Service's business model to the world of prediction APIs. In such a scenario, prediction API providers can leverage a Cloud provider's infrastructure to offer their inference service to the general public without having to worry about infrastructure acquisition and operation costs. Indeed, in the case of prediction APIs, self-hosting costs could be much higher than usual due to the fact that inference models, e.g., deep learning models, need specific hardware (e.g., graphical processing units) for an efficient execution. In such a context, trust is of great importance as the prediction API provider's most valuable asset, i.e., the inference model, is transferred to the Cloud provider. Thus, specific countermeasures should be designed to mitigate the possible attacks. In this paper, we analyze this scenario identifying the peculiar threat models. Then, we present a decentralized blockchain-based system, implemented on top of the popular Tendermint framework, that provides countermeasures to some of the main attacks. Numerical results, obtained executing deep neural network models, demonstrate that the overhead with respect to a centralized approach is negligible if compared with the advantages in terms of prevention of malicious behaviors.

To enable and support smart environments, a recent ICT trend promotes pushing computation from the remote Cloud as close to data sources as possible, resulting in the emergence of the Fog and Edge computing paradigms. Together with Cloud computing, they represent a stacked architecture, in which raw datasets are first pre-processed locally at the Edge and then vertically offloaded to the Fog and/or the Cloud. However, as hardware is becoming increasingly powerful, Edge devices are seen as candidates for offering data processing capabilities, able to pool and share computing resources to achieve better performance at a lower network latency—a pattern that can be also applied to Fog nodes. In these circumstances, it is important to enable efficient, intelligent, and balanced allocation of resources, as well as their further orchestration, in an elastic and transparent manner. To address such a requirement, this article proposes an OpenStack-based middleware platform through which resource containers at the Edge, Fog, and Cloud levels can be discovered, combined, and provisioned to end users and applications, thereby facilitating and orchestrating offloading processes. As demonstrated through a proof of concept on an intelligent surveillance system, by converging the Edge, Fog, and Cloud, the proposed architecture has the potential to enable faster data processing, as compared to processing at the Edge, Fog, or Cloud levels separately. This also allows architects to combine different offloading patterns in a flexible and fine-grained manner, thus providing new workload engineering patterns. Measurements demonstrated the effectiveness of such patterns, even outperforming edge clusters.

Nowadays, smart environments are becoming an integral part of our everyday lives. Objects are becoming smarter and the number of applications where they are involved increases day by day. In such a context, indoor localization is a key aspect for the development of smart services which are strictly related to the user position inside an environment. In this paper, we present a deep learning approach to estimate the indoor user location starting from its Wi-Fi fingerprint composed by those signals perceived in the environment. We show some experimental results that demonstrate the feasibility of the proposed approach.

The cover image, by Rustem Dautov et al., is based on the Research Article Metropolitan Intelligent Surveillance Systems for Urban Areas by Harnessing IoT and Edge Computing Paradigms, https://doi.org/10.1002/spe.2586. Photo Credit: Rustem Dautov.

Summary Recent technological advances led to the rapid and uncontrolled proliferation of intelligent surveillance systems (ISSs), serving to supervise urban areas. Driven by pressing public safety and security requirements, modern cities are being transformed into tangled cyber‐physical environments, consisting of numerous heterogeneous ISSs under different administrative domains with low or no capabilities for reuse and interaction. This isolated pattern renders itself unsustainable in city‐wide scenarios that typically require to aggregate, manage, and process multiple video streams continuously generated by distributed ISS sources. A coordinated approach is therefore required to enable an interoperable ISS for metropolitan areas, facilitating technological sustainability to prevent network bandwidth saturation. To meet these requirements, this paper combines several approaches and technologies, namely the Internet of Things, cloud computing, edge computing and big data, into a common framework to enable a unified approach to implementing an ISS at an urban scale, thus paving the way for the metropolitan intelligent surveillance system (MISS). The proposed solution aims to push data management and processing tasks as close to data sources as possible, thus increasing performance and security levels that are usually critical to surveillance systems. To demonstrate the feasibility and the effectiveness of this approach, the paper presents a case study based on a distributed ISS scenario in a crowded urban area, implemented on clustered edge devices that are able to off‐load tasks in a “horizontal” manner in the context of the developed MISS framework. As demonstrated by the initial experiments, the MISS prototype is able to obtain face recognition results 8 times faster compared with the traditional off‐loading pattern, where processing tasks are pushed “vertically” to the cloud.

Smart Cities are the perfect ecosystem where IoT technology could be applied. In particular, to allow a direct interaction with objets scattered over a city (e.g, parking slots, light posts, gates) a powerful and scalable architecture has to be envisioned. In this paper, using Bluetooth Low Energy as underlying technology, we extend the concept of Personal Area Networks to include smart city objects so that a user can easily interact with them in a geolocalized and real time manner. A prototype implementing a smart gate has been developed in our University Campus in order to demonstrate the feasibility of the proposed approach.

Cloud computing is becoming a relatively mature paradigm in the ICT landscape. In light of the growing appetite for resources and service levels on par with user expectations, multi-cloud scenarios are becoming the next frontier in the usage of distributed datacenters for private and hybrid Cloud scenarios. Application deployment in particular is a noteworthy feature to be evaluated as microservices become mainstream in adoption. Especially so when considered jointly with orchestration services; indeed OpenStack, as the most widely adopted Cloud middleware among the OpenSource community, features an orchestration subsystem, and may orchestrate the deployment of applications and services. In this work the authors will describe an architecture, developed within the H2020 BEACON project, for a standardized approach to orchestrated application deployment in multi-Cloud OpenStack- based setups, with TOSCA providing the specifications.

To kickstart the process of morphing Messina into a "smart" city, an explicit mission for the crowdfunded #SmartME project, it is essential to set up an infrastructure of smart devices embedding sensors and actuators, to be scattered all over the urban area. An horizontal framework coupled with the Fog computing approach, by moving logic toward the "extreme" edge of the Internet where data needs to be quickly elaborated, decisions made, and actions performed, is a suitable solution for data- intensive services with time-bound constraints as those usually required by citizens. This is especially true in the context of IoT and Smart City where thousands of smart objects, vehicles, mobiles, people interact to provide innovative services. We thus designed Stack4Things as an OpenStack-based framework spanning the Infrastructure-as-a-Service and Platform-as-a-Service layers. We present some of the core Stack4Things functionalities implementing a Fog computing approach towards a run- time "rewireable" Smart City paradigm, by outlining node management and contextualization mechanisms, also describing its usage in terms of already supported and developed verticals, as well as a specific example related to environmental data collection through #SmartME.

Cloud computing infrastructures provide services to a wide number of users whose behavior can deeply change at the occurrence of particular events. To correctly handle such situations a cloud infrastructure have to be reconfigured in a way that does not cause degradation in the overall performance. Otherwise, the quality of service specified in the service level agreement could be violated. To prevent such situations, the infrastructure could be organized as an autonomic system where self-adaptation and self-configuration techniques are implemented. Appropriate design choices become important in order not to fail in this goal. We propose a technique, based on a Petri net model and a specific analytical analysis approach, to represent Infrastructure-as-a-Service (IaaS) systems in the case in which the load conditions can suddenly change and reactive autonomic management techniques are applied to mitigate the consequences of the change. The model we propose is able to appropriately evaluate performance metrics in such critical situations making it suitable as a design tool for IaaS cloud systems.

Cloud monitoring is one important aspect for effective cloud management. Currently, cloud monitoring solutions can be classified into three groups: the ones not considering multiple layers or real time data analysis, the ones considering multiple layers but not real time data analysis, and the ones only considering real time data analysis. However, all these solutions fail to provide frameworks able to combine together monitoring in multiple layers and data stream analysis for detecting situations where multiple management actions are applicable in different layers of the cloud environment. This paper addresses this gap and proposes the Ceiloesper framework. Such a framework extends the OpenStack Ceilometer technology with Esper CEP and enables collection and analysis of information according to the principles defined in the 3-D cloud monitoring model, proposed in a previous work. The main contributions of this paper are: (i) the definition of the concept of Situation of Interest (SoI) leading to multiple management actions; (ii) the Ceiloesper architecture for a monitoring solution combining traditional monitoring elements with CEP; (iii) extensions to the Ceilometer OpenStack technology. We tested the Ceiloesper framework on a scenario based on the Wordpress application and the experimental results show its effectiveness.

Summary Performance and dependability evaluation plays a key role in the design of a broad range of systems, especially when strict requirements need to be met. This is particularly challenging in distributed contexts, where several components may interact among themselves by influencing each other. In this paper, we present an analytical method that allows the study of a class of systems where different operating conditions alternate by changing the stochastic behavior of the system components but still preserving the continuity of the performance and dependability quantities to investigate. The proposed solution technique, based on phase type distributions, Kronecker algebra, and ad‐hoc fitting algorithms, can be applied for the analytical evaluation of a wide class of distributed systems. Examples are provided to show the usefulness and the applicability of the methodology, characterizing and investigating different performance and dependability aspects of three distributed computing systems, that is, a connection‐oriented network, an Internet of Things application, and an Infrastructure‐as‐a‐Service Cloud. Copyright © 2014 John Wiley & Sons, Ltd.

Following the as-a-service philosophy, a cloud service provider offers computing utilities in the form of virtual resources instantiated on top of a physical infrastructure. In order to meet business requirements still providing high-quality services, performance evaluation needs to be carefully carried out with the aim of optimizing data center utilization and increasing user satisfaction. In this context, power efficiency plays a critical role pushing service providers towards the application of innovative green strategies. In this paper, we present an analytical framework, based on stochastic reward nets, that allows to evaluate different resource allocation policies in a green cloud. A use case is shown in order to illustrate the approach, modeling scattering and saturation allocation policies and comparing them to a purely physical data center scenario. A validation of the proposed model against the CloudSim framework is presented and several numerical results are provided, demonstrating the effectiveness of the approach as a powerful tool for a cloud service provider to perform well-informed decisions about the resource allocation policies to be enforced.

The transition to cloud computing offers a large number of benefits, such as lower capital costs and a highly agile environment. Yet, the development of software engineering practices has not kept pace with this change. Moreover, the design and runtime behavior of cloud based services and the underlying cloud infrastructure are largely decoupled from one another.This paper describes the innovative concepts being developed by CloudWave to utilize the principles of DevOps to create an execution analytics cloud infrastructure where, through the use of programmable monitoring and online data abstraction, much more relevant information for the optimization of the ecosystem is obtained. Required optimizations are subsequently negotiated between the applications and the cloud infrastructure to obtain coordinated adaption of the ecosystem. Additionally, the project is developing the technology for a Feedback Driven Development Standard Development Kit which will utilize the data gathered through execution analytics to supply developers with a powerful mechanism to shorten application development cycles.

Cloud data center management is a key problem due to the numerous and heterogeneous strategies that can be applied, ranging from the VM placement to the federation with other clouds. Performance evaluation of cloud computing infrastructures is required to predict and quantify the cost-benefit of a strategy portfolio and the corresponding quality of service (QoS) experienced by users. Such analyses are not feasible by simulation or on-the-field experimentation, due to the great number of parameters that have to be investigated. In this paper, we present an analytical model, based on stochastic reward nets (SRNs), that is both scalable to model systems composed of thousands of resources and flexible to represent different policies and cloud-specific strategies. Several performance metrics are defined and evaluated to analyze the behavior of a cloud data center: utilization, availability, waiting time, and responsiveness. A resiliency analysis is also provided to take into account load bursts. Finally, a general approach is presented that, starting from the concept of system capacity, can help system managers to opportunely set the data center parameters under different working conditions.

The key idea behind this work is the development of smart data centers able to monitor, control, and manage themselves through advanced analytics and management policies, collecting and analyzing real time data on their behavior. They also make adjustments to interdependent components across the physical infrastructure, in order to address changing business and technology needs. To this aim, we have designed and implemented a new framework for supporting green computing in the management of Cloud data centers. Through heterogeneous sensing devices deployed in the system, the framework is able to know the working state of the data center and to activate specific energy saving policies. We have evaluated the proposed solution through experiments on a real testbed implemented at the University of Messina. Experiments show interesting results thus proving real benefits deriving from the adoption of the proposed framework.

Smart environments represent a relatively uncharted ICT territory where plenty of sensor and actuator devices can be enrolled on-demand in order to realize high value-added services. A few application scenarios, such as Smart Cities, have already been explored. However, in order to finally enable such a paradigm, several issues have to be dealt with. In particular, from a developer perspective the high degree of heterogeneity for devices (ranging from cheap sensors to smart phones) could represent a hurdle for software design. In this paper, we present an innovative architecture that aims at providing a common reference platform for repurposing of devices i.e., reshaping their operational behavior for emergent and unforeseen requirements. Thanks to its modular and plugin based design, the proposed architecture is poised to ease implementation of both low-level (e.g., device discovery, code compilation, binary deployment) and high-level (e.g., service composition, data management) duties. We present the general architecture, then focusing on device-side aspects, while also providing two simple use cases that demonstrate the suitability of the proposed approach.

Cloud computing is a promising paradigm able to rationalize the use of hardware resources by means of virtualization. Virtualization allows to instantiate one or more virtual machines (VMs) on top of a single physical machine managed by a virtual machine monitor (VMM). Similarly to any other software, a VMM experiences aging and failures. Software rejuvenation is a proactive fault management technique that involves terminating an application, cleaning up the system internal state, and restarting it to prevent the occurrence of future failures. In this work, we propose a technique to model and evaluate the VMM aging process and to investigate the optimal rejuvenation policy that maximizes the VMM availability under variable workload conditions. Starting from dynamic reliability theory and adopting symbolic algebraic techniques, we investigate and compare existing time-based VMM rejuvenation policies. We also propose a time-based policy that adapts the rejuvenation timer to the VMM workload condition improving the system availability. The effectiveness of the proposed modeling technique is demonstrated through a numerical example based on a case study taken from the literature.

WS-BPEL language has become the industrial standard to design and orchestrate modular applications, formalizing service compositions and business relationships among providers and consumers. Once service level agreements (SLAs) among the parties are established, effective tools for evaluating appropriate measurements have to be developed to meet the requirements. However, the design of quality of service (QoS)-guaranteed composed Web services (WSes) still requires several efforts. This work aims at proposing a complete method to study the QoS of a composed WS at design time, i.e., when the process is specified by using WS-BPEL. Starting from the nonfunctional properties of the WS to compose, we propose a technique to derive non-Markovian stochastic Petri net (NMSPN) models from WS-BPEL processes, with the final goal of evaluating parameters such as the service time distribution and the service reliability. To demonstrate the effectiveness of the proposed method and to validate the obtained model, a nontrivial example implementing a travel agency flight reservation process, exposed as a synchronous composed WS, is investigated.

Cloud computing is a challenging technology that promises to strongly modify the way computing and storage resources will be accessed in the near future. Clouds may demand huge amount of energy if adequate management policies are not put in place. Optimization strategies are needed in order to allocate, migrate, consolidate virtual machines and manage the switch on/switch off period of a data center. In this paper, we present a modeling approach based on Stochastic reward nets to investigate the more convenient strategies to manage a federation of Clouds, having in mind the final goal to reduce the overall energy consumption. Several policies are presented and their impact is evaluated, thus contributing to a rational and efficient adoption of the Cloud computing paradigm.

Abstract In the last few years, remarkable efforts have been made to extend the Grid paradigm to commercial solutions. Business‐oriented grids call for effective Quality of Service strategies able to adapt to different user requirements and to address Service Level Agreements. Performance analysis and prediction with respect to different load conditions or management policies are required to define such strategies. However, the highly distributed nature of Grid systems and the presence of distinct administrative domains make it difficult to carry out performance estimations. In fact, several parameters are involved and the autonomy of each site could make it complex to set them in a proper way. In this paper, we present a non‐Markovian Stochastic Petri Net methodology that allows to conduct performance analysis of Grid systems focusing on aspects related to the Virtual Organization as a whole. In particular, different job allocation techniques can be evaluated with respect to both user and provider points‐of‐view. The influence of different information update policies on the accuracy of the allocation schemes can also be investigated, highlighting the costs/benefits in terms of job waiting time, service availability, and system utilization. The proposed methodology is designed to be as general as possible and it can be applied to analyze a gLite Grid infrastructure taken as case study. Copyright © 2011 John Wiley & Sons, Ltd.

Service Oriented Architecture (SOA) is the most important and effective software paradigm to design Internet-based services. Using the SOA technology, value-added services can be easily deployed as a combination of existing Web services. In this context, WS-BPEL language has become the SOA industrial standard. To allow services to be composed, business relationships between providers and consumers have to be adequately managed. This implies that a formal definition of Quality of Service (QoS) is agreed and that effective tools for its measurement have to be developed. However, the design of QoS guaranteed composed Web services still requires several efforts due to the highly distributed nature of such software applications. This work aims at proposing a methodology to evaluate Web service performance at the earliest design phase. We present a novel technique to translate WS-BPEL processes into non-Markovian stochastic Petri nets with the final goal to evaluate parameters such as service time distribution and service reliability. The obtained model can be numerically solved through automatic tools, allowing to investigate the service behavior under different operating conditions and thus helping software engineers to develop QoS-guaranteed software solutions.

Grid Computing supports the shared and coordinated use of several resources in dynamic Virtual Organizations. In the last few years, it is evolving into a business-innovating technology that is driving commercial adoption. Such a new scenario calls for powerful strategies able to guarantee stringent QoS requirements in order to meet Service Level Agreements (SLAs) between customers and providers. For this reason, it is necessary to analyze and predict performance with respect to different load conditions or management strategies. In this paper, we present a methodology to analyze performance in gLite Grids through the use of Generalized Stochastic Petri Nets (GSPNs). We introduce a cluster-level model of a typical gLite site taking into account the coexistence between normal and MPI-based jobs. We investigate the influence of different strategies (e.g., scheduling) on the performance of the whole site, highlighting aspects related to both customer and provider point of views. We also provide a business-oriented performance analysis introducing two different SLA typologies and highlighting how the site configuration may influence the expected profit of the service provider.

Wireless networks are a very challenging communication technology since their ability to be set everywhere and whenever. Among the several types of wireless systems, a new class of networks is gradually emerging: Wireless Mesh Networks (WMNs). A WMN is a distributed communication infrastructure organized in a mesh topology, which handles multi-hops connections and is capable of provide dependable services by dynamically updating and optimizing communications. This paper presents a new cross-layer architecture for supporting QoS in WMNs. It integrates the DiffServ paradigm with the admission control signaling and a multipath routing to route QoS traffic along reserved paths, while making use of alternative paths for Best Effort load. Performance measurements, based on simulative techniques are carried out to test the reliability of the proposed system.

Starting from 1990s and until now, Grid computing has been mainly used in scientific laboratories. Only in the last few years, it is evolving into a business-innovating technology that is driving commercial adoption. In this paper, we describe GridVideo, a Grid-based multimedia application for the distributed tailoring and streaming of media files. The objective is to show, starting from a real experience, how Grid technologies can be used for the development of nonscientific applications. Relevant performance aspects are analyzed, regarding both user-oriented (in terms of responsiveness) and provider-oriented (in terms of system efficiency) requirements. Different multimedia data dissemination strategies have been analyzed and an innovative technique, based on the Fibonacci series, is proposed. To respond to the stringent quality-of-service (QoS) requirements, typical of soft real-time applications, a reservation-based architecture is presented. Such architecture is able to manage the Grid resource allocation, thus enabling the provisioning of advanced services with different QoS levels. Technical and practical problems encountered during the development are discussed, and a thorough performance evaluation of the developed prototype is presented.

Wireless ad hoc networks experienced a great diffusion thanks to their simple deployment everywhere and whenever needed. Already used for entertainment applications (e.g., instant messaging) or for specific domains (e.g., military) we think such technology is mature enough to support also real-time services provision.Real time applications pose several challenges, in particular issues such as limited bandwidth, unreliable channels, topology evolutions and power consumption make quality of service (QoS) management a mandatory task to be addressed. In this paper, we propose a general QoS system that makes use of differentiated services strategies by exploiting the concept of virtual backbones. Our key idea is to use an auto-configured virtual backbone to set up a differentiated service area in order to opportunely manage QoS flows. Also, thanks to a new distributed call admission control algorithm, it avoids the arise of overloading situations. Performance measurements, based on simulative techniques and carried out to test the feasibility of the proposed system, are finally presented.

Wireless sensor networks are a new emerging area where swarm intelligence can be applied with interesting implications. In fact, a strong analogy between unicellular organism colonies and wireless sensor networks can be emphasized: a sensor network can be viewed as a "colony" of simple, scarce resource nodes that, autonomously, are able only to perform simple tasks, but all together can accomplish very complex problems. In this paper we propose a routing protocol with interesting properties: self organization, fault tolerance and environmental adaptation. The proposed protocol was inspired by the well known behavior (in artificial life studies) of "Slime Mold". Such colony of unicellular organisms organizes itself in clusters by pheromone generation and evaporation mechanisms. In a similar manner our protocol manages the data traffic in clusters towards the sink nodes using the gradient concept and reaching high levels of autonomy. We analyze the proposed protocol to examine the performances and the adaptation properties using simulation techniques.

Abstract Very popular mobile devices such as cellular phones, laptops, and handheld and tablet PCs are becoming more and more powerful. One of today's trends is to access multimedia services from this sort of device. Their low capacity in terms of storage, and their high inclination to wireless connection, suggest the use of streaming techniques instead of media download. Mobile devices are, at present, dramatically unhomogeneous for visualization and audio reproduction characteristics. Consequently, today multimedia delivery systems are very customized for some device profiles or for wireless connection types. Transcoding techniques are used in order to adapt the same media source to the multiple hardware profiles. Client‐based transcoding techniques show their weakness in high energy consumption, and server‐based transcoding techniques suffer from scalability problems. In this paper we propose a scalable distributed multimedia server based on the Grid computing paradigm, which can adapt media content to the device profile and/or to the connection link. Copyright © 2005 John Wiley & Sons, Ltd.

Mobile networks are requested to provide several added value services with different QoS levels. An efficient call admission control (CAC) is one of the strategic components that a network management policy requires, which defines how to allocate network resources to roaming and new users. We assume a hierarchical network architecture constituted of several cellular IP regions which may communicate each other accessing the Internet through a Mobile IP protocol. In such scenario, we demonstrate the importance of CAC not only in the ingress wireless points, but in the wired network too. An integrated approach to CAC is then proposed whose advantages are shown through simulation analysis.

Resource reservation in wired networks is a relatively mature area, with some available solutions already on the market. Due to the increasingly growing diffusion of mobile terminals and services, the study of the issues concerning the management of mobility in IP-based networks is also arising increasing interest. The combination of both issues is however still at its initial definition. This work is based on the idea of combining mobile architectures with the available techniques for resource reservation. In particular, we have created a cellular IP-based architecture, which can assure some agreed levels of QoS (for instance, in terms of data throughput) according to the IP layer. We present the proposed architecture, describe its implementation and show some experimental results for its validation.

This paper wishes to investigate the converging field of mobile and Grid computing by defining an architecture for the provision of Grid services, which Is based on standards, robust and useful across application domains. We propose using the mobile agent paradigm in order to develop a middleware layer that takes care of all the details to allow mobile users to access distributed resources in a transparent, secure and effective way. Our purpose Is also that of Identifying the environmental situations In which such paradigm should be preferred or adopted in conjunction with more traditional communication paradigms (i.e. client/server, Remote Evaluation). For this purpose, we provide an experimental and analytical evaluation of the Client-Server, Remote Evaluation and Mobile Agent communication paradigms.

Wireless devices are becoming very popular and powerful to be commonly adapted to access distributed services. More and more sophisticated applications are being developed and a very promising market is quickly being established where mobile users might access multimedia data while roaming from a cell to another, anytime and everywhere. Such scenario requires a strong infrastructure in order to adequately manage all the different issues related with high-level service provisioning. QoS represents one of the most crucial issues as it involves many different aspects and directly impacts the user satisfaction. In order to achieve this target, we propose an architecture, which allows mobile devices to access advanced services available in the wired part of the network. This architecture, based on mobile agents technology, assumes the presence of a VOD virtual server. The strategy adopted is to hide all the basic mechanisms inside a middleware layer that provides necessary interfaces to allow a simple interaction between the user on the one side and the network and the distributed services on the other. The overall architecture, that is currently being implemented and tested, has been used as an experimental environment to provide mobile users with a new service to access MPEG-4 flows.