Bringing intelligent decision-making to industrial facilities involves coordinating a complex ecosystem of applications with different performance requirements that often depend on each other. A promising approach to managing this complexity is leveraging loosely coupled software architectures, such as microservices, with distributed infrastructures, such as the Edge-to-Cloud Continuum. The order in which microservices are provisioned in the Edge-to-Cloud Continuum can significantly impact outcomes when distributing tasks to agents. Poor provisioning orderings can lead to non-critical microservices overburdening servers with lower latency while leaving latencysensitive microservices with no option but to run on more distant servers. To address this challenge, we introduce a novel provisioning ordering mechanism within a Multi-Armed Bandit (MAB) algorithm, helping to conquer the problem. Simulations based on real-world specifications indicate a $\mathbf{3 8 \%}$ reduction in latency violations and up to 6% decrease in power consumption compared to an unordered approach.

Power-performance efficiency has become one of the most critical issues in evolving High-Performance Computing systems (HPC) towards Exaflops. Thread-level parallelism (TLP) exploitation, dynamic voltage and frequency scaling (DVFS), and uncore frequency scaling (UFS) are methods widely applied to better balance the power consumption and performance improvements of parallel applications. However, selecting ideal combinations of these knobs for every application is challenging due to the massive number of possible solutions, as there is no unique combination that delivers at the same time the best performance and the lowest power consumption. Given that, we propose HPC-PPO (power-performance optimizer), a multi-objective optimization strategy driven by an artificial neural network that leverages hardware and software features of parallel applications to predict Pareto-efficient configurations of TLP degree, DVFS, and UFS that optimize the balance between power and performance. When validating HPC-PPO on three multicore processors with twenty-five applications, we show that HPC-PPO can predict combinations very close to the best ones found by an exhaustive search. We also show that the Pareto-efficient configurations predicted by HPC-PPO improve parallel applications' performance by 30.7% while spending 23.9% less power when compared to state-of-the-art strategies.

The pursuit of self-driving networks is increasing pressure on adopting intelligent, edge-based networking services. However, deploying autonomous network models within operational and large-scale infrastructures entails substantial risks that require rigorous verification and validation procedures. In this context, the application of a Network Digital Twin (NDT) is emerging as a viable approach towards intelligent network decision-making based on high-fidelity models built upon digital representations of physical network devices (i.e., Digital Twins). In this paper, we take the first steps towards efficiently provisioning NDT models. To that end, we introduce the Digital Twin Network Provisioning Problem (DigiNet), which encompasses the optimal placement of NDT models and the efficient collection of telemetry data for synchronizing NDT models with their physical counterparts. We theoretically formalize DigiNet as a Mixed-Integer Linear Programming (MILP) model and present a polynomial-time heuristic. Our results show that DigiNet outperforms baseline approaches by up to 10x regarding the number of NDT models provisioned.

Data plane programmability is redesigning the way we manage and operate forwarding devices. However, most of the algorithmic decisions performed by data planes are still deterministic and control-plane dependent. We argue that it is possible to break this dependency and make the data plane intelligent, so that it can learn the infrastructure state autonomously. Despite existing efforts to make data planes intelligent, little has been done to design unsupervised ML algorithms that fit the architectural constraints of programmable devices. Executing such approaches in the data plane has the potential to reduce the overall decision-making time, thus meeting packet processing deadlines (which are in the order of nanoseconds). In this paper, we propose Spinner, the first effort to deliver an unsupervised Machine Learning (ML) approach entirely in programmable devices. Spinner is a flow clustering algorithm designed to fit existing architectural constraints of SmartNICs, and that can reach line rate for most packet sizes with complexity O(k). To demonstrate the potential behind in-network clustering, we prototype and deploy Spinner in a programmable testbed and use it to enhance Explicit Congestion Notifications (ECN) at the server side. Spinner-enhanced TCP provides up to 2x higher throughput when comparing to de-facto TCP implementations.

Federations between Edge Computing infrastructure providers represent a promising approach for improving the applications' Quality of Service (QoS) and the infrastructure's resource usage. At the same time, federated edges impose particular provisioning challenges, as data protection policies implemented by certain providers within a federation may conflict with the privacy requirements of services carrying out sensitive information (e.g., databases). In addition, the popularization of complex software architectures (e.g., composite applications) sets strict latency requirements that narrow the provisioning possibilities even further. Previous research efforts targeting federated edges have focused either on coupling with end-user performance requirements (e.g., latency and privacy) or on satisfying infrastructure providers' objectives (e.g., power consumption reduction), but none on balancing both. This paper presents Thea, a novel approach for provisioning composite applications on federated edges which optimizes applications' latency and privacy while reducing the infrastructure's power consumption. Simulated experiments show that Thea can achieve near-optimal results, reducing application latency and privacy issues by 50% and 42.11% and the infrastructure's power consumption by 18.95% compared to state-of-the-art approaches.

Container-based virtualization represents a low-overhead and easy-to-manage alternative to virtual machines.On the other hand, containers are more prone to performance interference and unpredictability.Consequently, there is growing interest in predicting and avoiding performance issues in containerized environments.Existing solutions tackle this challenge through proactive elasticity mechanisms based on workload variation predictions.Although this approach may yield satisfactory results in some scenarios, external factors such as resource contention can cause performance losses regardless of workload variations.This paper presents Flavor, a machine-learning-based system for predicting and avoiding performance issues in containerized applications.Rather than relying on workload variation prediction as existing approaches, Flavor predicts application-level metrics (e.g., query latency and throughput) through a deep neural network implemented using Tensorflow and scales applications accordingly.We evaluate Flavor by comparing it against a state-ofthe-art resource scaling approach that relies solely on workload prediction.Our results show that Flavor can predict performance deviations effectively while assisting operators to wisely scale their services by increasing/decreasing the number of application containers to avoid performance issues and resource underutilization.

The popularity of mobile devices has increased significantly, and nowadays they are used for the most diverse purposes like accessing the Internet or helping on business matters. Such popularity emerged as a consequence of the compatibility of these devices with a large variety of applications. However, the complexity of these applications boosted the demand for computational resources on mobile devices. Code Offloading is a solution that aims to mitigate this problem by reducing the use of resources and battery on mobile devices by sending parts of applications to be processed in the cloud. In this sense, this paper presents an evaluation of a transparent code offloading technique, where no modification in the application source code is required to allow the smartphone to send parts of the application to be processed in the cloud. We used a face detection application for the evaluation. Results showed the technique can improve applications performance in some scenarios, achieving speed-up of 12x in the best case.

The integration of disciplines in higher technology courses is a significant challenge. Sometimes the student can not visualize the correlations between different knowledge in the direction of a complete formation. This paper shows an integrated professional practice that relies on an innovation discipline to discover and prospect new technological products. From this, technical disciplines can focus on developing such products, having as supporting the knowledge and capacities determined in their curricula. The results show that students were able to develop products that aligned with local and regional problems, as well as open up a possibility for the creation of startups based on the developed products.

Teachers and psychologists report that students may suffer from multiple psychological issues such as lack of interest and high levels of stress during the development of final course projects. In this context, approaches such as gamification arise with the proposal of improving students motivation by bringing games elements to school. However, employing gamification into classroom is not a trivial task since, if not managed properly, students may lose their focus. In this paper we present GAMED, an assessment methodology that introduces sistematic steps to improve students engagement through gamification. We used GAMED in a class with high school students during two semesters and the results showed that it can improve aspects such as motivation, engagement, and teamwork.

In order to meet the increasing performance demand of applications, the amount of cores in a single chip package has been increasing. However, the heat has been rising at a higher scale, which accelerates the aging process in modern processors. Therefore, wisely balancing the use of resources is important to extend its longevity. Frequency performance stagnates after a certain amount of concurrent threads starts executing. In such cases, the only result is a temperature rise that directly influences the aging process, reducing the processor lifetime. This unbalance between threads can be originated from many factors, which includes the way threads communicate and synchronize. Considering that those characteristics are related to the Parallel Programming Interface (PPI) used to parallelize the application, this work proposes to evaluate three widely used PPIs executing on an embedded multicore. We show that, depending on the characteristic of the application, by only switching from one PPI to another, it is possible to reduce the effects of aging. For that, we have developed a model based on the Arrhenius equation. We show that OpenMP has a lower impact on the processor aging for memory-bound applications: up to 38% and 68% lower than PThreads and MPI, respectively. On the other hand, PThreads presents the lowest impact on the processor aging for CPU-bound applications.

The massive power consumption of data centers has been a recurring concern in current research. In cloud environments, lots of methods are being adopted that aim for energy efficiency. However, although such methods enable the decrease in power consumption, they regularly affect application performance. In this paper, we present a multilevel resource allocation approach towards dynamic network bandwidth at the physical substrate, managing different power-saving states and workload allocation at the cloud infrastructure at the same time employ virtual machine allocation and selection policies at the cloud platform. In order to evaluate our approach, tests were carried out in a simulated environment using scale-out application on a dynamic cloud infrastructure. Results showed that our proposal presents a better balance regarding a more energy-efficient data center with a smaller impact on application performance when compared with other works discussed in the literature.

Virtualization has become a key technology that provides several advantages (e.g., flexibility, migration, isolation) for a plethora of computing infrastructures. However, traditional virtualization models are not suitable for embedded IoT platforms due to the virtualization layer verhead. New virtualization proposals such as container-based approaches arise as an option where performance is not impacted. However, when working on general-purpose embedded platforms, some studies have demonstrated that applications on container-based virtualization on embedded devices present considerable performance overhead. Since most performance evaluations on platforms using containers were run on servers, this study expands the testbed scenario by analyzing several metrics that measure the overhead of container-based virtualization layer on embedded IoT devices. Results demonstrated improvements up to 23% in terms of performance and up to 32% in terms of EDP.

Embedded systems are becoming more popular in several sectors of society by performing a broad range of tasks. In this context, there is a concern about improving the trade-off between performance and energy savings since they are usually battery-dependent. However, it is not a trivial task since some embedded devices are developed with strict hardware constraints. In this sense, we present an operating system level tool that groups processes dynamically on resources. Our tool manages different types of processes, and through isolation characteristics, provides a better utilization of resources. The results show that our tool can improve the trade-off between performance and power saving of embedded systems in up to 15%.

Energy saving management in multi-core embedded environments has been a challenge for designers. To achieve energy efficiency, most studies consider dynamic frequency scaling on one hardware component only, such as processor or memory - which will most likely also affect performance. This work proposes the use of frequency scaling considering the three most important hardware components altogether: processors, L2 cache, and RAM; seeking for the best set of frequencies for each one of them to improve the Energy-Delay Product (EDP), depending on the application's behavior. Therefore, this work addresses multidimensional frequency scaling for multi-core embedded systems. By evaluating different frequency levels, we show that the EDP can be improved in up to 46.4% when compared to the standard way that the frequencies are configured.