Applications such as autonomous vehicles, virtual reality, augmented reality, and heavy machine learning-based applications are becoming popular and demanding more flexible deployment environments. The computing continuum, a hierarchical hybrid infrastructure comprehending user devices (smartphones, sensors, laptops, etc.), edge data centers, and cloud platforms, offers a wide range of deployment possibilities with a full range of varying computing resources. To take full advantage of such infrastructure, application development is faced with many challenges, the most important being the implementation of a transparent and generalized mechanism for code offloading and mobility throughout the continuum. To tackle such issues, this paper presents the Self-Distributing Systems (SDS) framework, a self-distribution framework that supports generalized code-offloading capabilities at the application level with a machine learning agent for deciding where to place components and a component-based model to enable seamless distribution of an application's components at runtime. We describe the framework, show its applicability in different application scenarios, and report our preliminary results. We conclude the paper with a list of challenges and invite the systems community to join the effort to further investigate them.

Multiway spatial joins are a commonly occurring and fundamental type of query for spatial data processing. This article presents models and algorithms to schedule this type of query in distributed database systems while attempting to strike a balance between makespan and communication costs. We propose three algorithms based on combinatorial optimization methods: the well-known linear relaxation technique of rounding a solution generated by linear programming (LP), a more sophisticated Lagrangian Relaxation method (LR), as well as a greedy heuristic (GR) for baseline comparison. Our evaluation shows that a schedule built using GR consumes, on average, 22% more processing and communication resources than a more elaborate schedule constructed via the LR method, when scheduling a query for 64 machines. The schedule provided by LR is also, on average, an order of magnitude closer to the optimal schedule for a query compared to GR. We show that scheduling Gigabyte-size multiway queries before execution can reduce its processing time by an order of magnitude compared to state-of-the-art frameworks for spatial data processing that do not have this capability, and can significantly reduce the amount of shuffled data in the network.

Contemporary deployment environments are volatile, with conditions that are often hard to predict in advance, demanding solutions that are able to learn how best to design a system at runtime from a set of available alternatives. While the self-adaptive systems community has devoted significant attention to online learning, there is less research specifically directed towards learning for open-ended architectural adaptation - where individual components represent alternatives that can be added and removed dynamically. In this paper we present the Emergent Web Server (EWS), an architecture-based adaptive web server with 42 unique compositions of alternative components that present different utility when subjected to different workload patterns. This artefact allows the exploration of online learning techniques that are specifically able to consider the composition of logic that comprises a given system, and how each piece of logic contributes to overall utility. It also allows the user to add new components at runtime (and so produce new composition options), and to remove existing components; both are likely to occur in systems where developers (or automated code generators) deploy new code on a continuous basis and identify code which has never performed well. Our exemplar bundles together a fully-functional web server, a number of pre-packaged online learning approaches, and utilities to integrate, evaluate, and compare new online learning approaches.

Smart Cities combine advances in Internet of Things, Big Data, Social Networks, and Cloud Computing technologies with the demand for cyber–physical applications in areas of public interest, such as Health, Public Safety, and Mobility. The end goal is to leverage the use of city resources to improve the quality of life of its citizens. Achieving this goal, however, requires advanced support for the development and operation of applications in a complex and dynamic environment. Middleware platforms can provide an integrated infrastructure that enables solutions for smart cities by combining heterogeneous city devices and providing unified, high-level facilities for the development of applications and services. Although several smart city platforms have been proposed in the literature, there are still open research and development challenges related to their scalability, maintainability, interoperability, and reuse in the context of different cities, to name a few. Moreover, available platforms lack extensive scientific validation, which hinders a comparative analysis of their applicability. Aiming to close this gap, we propose InterSCity, a microservices-based, open-source, smart city platform that enables the collaborative development of large-scale systems, applications, and services for the cities of the future, contributing to turn them into truly smart cyber–physical environments. In this paper, we present the architecture of the InterSCity platform, followed by a comprehensive set of experiments that evaluate its scalability. The experiments were conducted using a smart city simulator to generate realistic workloads used to assess the platform in extreme conditions. The experimental results demonstrate that the platform can scale horizontally to handle the highly dynamic demands of a large smart city while maintaining low response times. The experiments also show the effectiveness of the technique used to generate synthetic workloads.

Offering fluid multi-device interactions to users while protecting their privacy largely remains an ongoing challenge. Existing approaches typically use a peer-to-peer design and flood session information over the network, resulting in costly and often unpractical solutions. In this paper, we propose Sprinkler, a decentralized probabilistic dissemination protocol that uses a gossip-based learning algorithm to intelligently propagate session information to devices a user is most likely to use next. Our solution allows designers to efficiently trade off network costs for fluidity, and is for instance able to reduce network costs by up to 80% against a flooding strategy while maintaining a fluid user experience.

Middleware was introduced to facilitate the development of sophisticated applications based on a uniform methodology and industry standards. However, early research and practice suggested that no one-size-fits-all approach was suitable for all application domains and scenarios. This gave rise to industry initiatives to standardize domain-specific middleware services and profiles, as well as research efforts on configurable, reflective, and adaptive middleware. The industry's approach led to easy deployment, although with a level of flexibility limited by the extent of existing profiles. The approach of the research community, on the other hand, enabled high flexibility, allowing any middleware configuration to be defined. Nevertheless, creating sound configurations using this approach is a challenging task, limiting the target audience to expert engineers. As a consequence, both initiatives do not scale with the current proliferation of specialized application domains. In this paper, we target this problem with an approach that leverages model-driven engineering for the construction of domain-specific middleware platforms. A set of high-level, yet expressive, building blocks is defined in the form of a metamodel, which is used to create models that specify the desired middleware configuration. We argue that this approach enables the rapid development of middleware platforms to match the proliferation of application domains, at the same time as it does not require per-application middleware construction or even highly skilled middleware engineers. We present the current state of our research and discuss research directions to fully realize the approach.

Introductory programming courses entail students' high failure and dropout rates. In an effort to tackle this problem, we carried out a qualitative study aiming to shed some light on the programming phase that is most challenging for students, in order to elicit the specific difficulties they experience while learning to program. In doing so, distinctive cognitive abilities, differentiating subjects in terms of the way they handle programming tasks, were detected. Such aptitudes are represented in three groups of students: those who learn easily, those who never seem to fully grasp what programming requires despite true effort, and those who experience a sudden insight, making them leap from a point were they had difficulties to another where they overcome them. By interviewing teachers and students, abstraction and sequencing elaboration were found to be the two core skills for programming. These results impelled us to consider the mental models' approach, concluding that there are very specific cognitive functions that are more favorable to learn programming and that are fostered by more adequate schemas of representing reality. Some conclusions involving Problem-based learning as a fit teaching methodology to overcome students' difficulties are also presented.

SUMMARY Opportunistic grid computing middleware has as a main concern the need to preserve the performance of the local applications running on machines that donate resources to the grid. This concern, together with the fact that it happens in an extremely dynamic environment, causes the adoption of a treatment based on thetextitbest‐effort principle for grid applications. This means that efficient application management schemes are usually not employed, which results in less than optimal performance as grid applications often need to be restarted due to (often temporary) resource claims by local user applications. This paper presents a method to improve the performance of grid applications, taking into account resource usage profiles for local applications, trying to identify when such resource claims are temporary and avoiding costly actions such as the migration of grid tasks. The approach is proposed as an extension to the InteGrade middleware and its evaluation shows promising results for the efficient management of grid applications. Copyright © 2011 John Wiley & Sons, Ltd.

The introductory undergraduate course on Algorithms and Computer Programming, commonly known as CS1, has always presented a challenge when considering student failure and drop out rates. Despite this, it is acknowledged that this is a foundational course for a large part of the CS curriculum. In this paper we present the results of a project that combines the use of mobile, pen-based, computing technology and Problem-Based Learning in the redesign of an introductory computer programming course. The course redesign focused on the integrated use of tablet PCs to assist in the several activities involved in the use of the PBL method in the classroom. The results show a promising future for the methodology, also pointing to the need for some important adaptations in order to make its use more effective to teach and learn this particular discipline.

Abstract The message passing interface (MPI) is a popular programming model for parallel applications. Support for MPI in grid middleware is important for the widespread use of grids for parallel programming. This enables existing parallel applications to be executed on large‐scale grids, as opposed to being restricted to local clusters. In the specific case of opportunistic grids, the use of idle computing power from non‐dedicated computers further adds to the range of resources that can be used. In this paper we present MPICH‐IG, an implementation of the MPI‐2 standard on top of the InteGrade grid middleware. Existing MPI applications can run unmodified, while taking advantage of the InteGrade scheduler to harvest the available computing power from the grid. In addition, fault‐tolerance of MPI applications is achieved through a checkpointing mechanism, which allows applications to be resumed after failures of particular grid nodes. Copyright © 2009 John Wiley & Sons, Ltd.

In addition to an untuned performance, inefficient resource management in hinders any attempt to offer Quality of in scavenging grids. In this case, Best-Effort mechanisms are synonym to unreliability. Evidently it would be impossible, with undedicated resources, to offer any deterministic warranty (if based on a classic reservation of resources) to service access. Although the usage of scavenging grids for real-time execution remains a challenge, new Service Qualities may be proposed to compatiblize applications' preferences versus system oscillations.

Reflection has found its way into middleware platforms as a principled technique to provide flexible infrastructures that can dynamically adapt to meet new or changing requirements. In this context, there is a common belief that the use of such a technique incurs unacceptable overheads. In this paper, we evaluate the performance of reflective middleware, comparing it with non-reflective approaches and with requirements of an important class of applications. The aim is to show that, although the associated overhead cannot be ignored, it is well within acceptable limits. The work is done in the context of a particular reflective middleware platform, the meta-ORB python prototype of open ORB. In this context, we also demonstrate that the integration of meta-information management and reflection does not compromise performance, at the same time as it allows a uniform programming model.