INFORUM 2017: 9TH INFORUM - INFORMATICS SYMPOSIUM [THE ORGINAL TITLE IS IN PORTUGESE: 9º INFORUM - SIMPóSIO DE INFORMáTICA]
PROGRAM FOR FRIDAY, OCTOBER 13TH
Days:
previous day
all days

View: session overviewtalk overview

09:30-10:30 Session 6A: Sistemas Embebidos e de Tempo Real - SETR
Chair:
José Rufino (University of Lisbon, Faculty of Sciences, LaSIGE, Portugal)
Location: Salão Nobre
09:30
João De Sousa Alves (University of Lisbon, Faculty of Sciences, Portugal)
José Rufino (University of Lisbon, Faculty of Sciences, LaSIGE, Portugal)
Comparing the inaccessibility characteristics of CAN and CAN FD protocols
SPEAKER: unknown

ABSTRACT. The Controller Area Network (CAN) protocol, originally designed more than two decades ago, has been widely used in multiple domains, including industrial control and automotive applications. The aerospace industry, traditionally conservative and very cautious in the adoption of new technologies, has started using CAN in the beginning of the XXI century.

Since its definition, the original CAN protocol exhibit two important limitations: low data transmission rates (only up to 1 Mbps) and small data frame payload sizes (8 bytes, maximum).

An effort was made to overcome those limitations and a new specification, known as CAN FD (CAN Flexible Data rate), was issued. However, after having studied and analysed the timing and safety characteristics of the new specification, several technical issues arise. Does CAN FD significantly improve network timeliness and data transfer bandwidth? Does CAN FD significantly improve network robustness in the presence of errors, including reducing the protocol recovery times and (non-intrusive) observability of the protocol state? Has the protocol become simpler, more effective and easier to integrate into core chip designs?

This paper aims to contribute to the provisioning of answers to those questions, showing that CAN FD, although showing an improvement of timeliness in the absence of errors, continues to exhibit (almost) the same shortcomings of the original protocol with respect to its use in the design of highly reliable real-time embedded systems.

09:50
Daniel Mendes (University of Coimbra, Portugal)
Jose Nunes (Instituto Politécnico de Coimbra, Portugal)
Sérgio Patrão (University of Coimbra, Portugal)
Naghmeh Ivaki (University of Coimbra, Portugal)
Pedro Amaro (Instituto Politécnico de Coimbra, Portugal)
João Carlos Cunha (Instituto Politécnico de Coimbra, Portugal)
Assessing the Robustness of a Quadcopter's Flight Controller to Sensor Failures
SPEAKER: unknown

ABSTRACT. The rapidly growing range of Unmanned Aerial Vehicle (UAV) applications provides the basis for an increasing number of commercial and scientific solutions. One popular deployment of UAVs is on multi-rotor helicopters, namely quadcopters, which have been documented to be used in very diverse activities such as critical search and rescue, precision agriculture, crowd surveillance or hobby photography. Independently from their mission, the expanding use of quadcopters raises real concerns related with the possibility of human injury due to malfunction. For example, a failure in an inertial sensor may cause an erroneous reaction of the flight controller, making a hazardous maneuver.

The performance and safe operation of a flight controller is heavily dependent on the perception of the environment, and as such it relies on a series of redundant and complementary sensing devices (e.g., accelerometer, gyroscope, barometer, and GPS), whose data is combined for correcting the deficiencies of the individual sensors and calculating accurate information concerning position, orientation and acceleration.

In this paper we assessed how effective is the combination of data from sensors (sensor fusion) and the flight control algorithm to tolerate failures in the sensors. To this end, we injected a number of faults in the sensors of a quadcopter using the ArduPilot flight controller platform. The quadcopter is programmed with several autonomous flight missions, and its behavior analyzed within a simulation environment.

10:10
Luís Nóbrega (Instituto de Telecomunicações Aveiro, Portugal)
Pedro Gonçalves (ESTGA/IT - Universidade de Aveiro, Portugal)
Paulo Pedreiras (DETI/IT - Universidade de Aveiro, Portugal)
Rui Morais (Instituto de Telecomunicações Aveiro, Portugal)
André Temprilho (Instituto de Telecomunicações Aveiro, Portugal)
SheepIT: Automated Vineyard Weeding Control System

ABSTRACT. Controlling the growth of weeds in vineyards is an actual and important issue for the viticulture sector. Existing methods are ag- gressive to the environment and plants, costly and even potentially dan- gerous to the human health. Using animals, as sheep, for weeding, is an ancient method that does not suer from these problems. However, an- imals tend to feed from the vine branches and fruits, thus constituting themselves a threat to the wine production, and require costly human supervision. The SheepIT project arose to mitigate these disadvantages, thought the implementation of an IoT-based architecture that monitors and controls sheep's behavior, ensuring that they only feed from weeds and stay conned to designated areas, without requiring constant human supervision.

10:15
Luis Marques (ISEC/IPC, Portugal)
Luis Almeida (Fac. de Eng. da Universidade do Porto, Portugal)
Paulo Pedreiras (IT, Universidade de Aveiro, Portugal)
Reactive Error Recovery in Time-Triggered Networks using On-line Traffic Scheduling
SPEAKER: unknown

ABSTRACT. Time-Triggered (TT) networks are a common choice in safety critical systems, giv-en their high predictability. However, their typical static traffic scheduling precludes bandwidth efficient handling of sporadic events, such as error retransmissions. We address this limitation resorting to networks that use on-line traffic scheduling, such as those based on the Flexible Time-Triggered (FTT) paradigm, and particu-larly FTT-CAN. On-line traffic scheduling supports prompt error recovery while avoiding bandwidth overprovisioning of typical recovery approaches in TT net-works that reserve extra communication slots in static traffic schedules. The pro-posed approach requires significantly less bandwidth while providing timely recov-ery.

10:20
João De Sousa Alves (University of Lisbon, Faculty of Sciences, Portugal)
José Rufino (University of Lisbon, Faculty of Sciences, LaSIGE, Portugal)
On the (non-intrusive) observability of the CAN FD protocol
SPEAKER: unknown

ABSTRACT. The Controller Area Network (CAN) protocol has been extensively used in many application domains, including industrial control, appliances, medical, and transportation. The last sector includes manned and unmanned vehicles in land, maritime and aerospace.

The definition of the CAN with Flexible Data rate (CAN FD) specification, currently under normalisation, enhances the original CAN protocol in two ways: it extends from 8 to 64 bytes the maximum payload size of a data frame; it enhances the bus signalling rate, while maintaining the determinism of node network access arbitration.

The CAN FD data frame includes in its header an Error State Indicator (ESI) flag. Thus, one fundamental question is whether or not the ESI flag will be useful for building highly reliable distributed real-time embedded systems based on the CAN FD protocol?

This communication formulates the problem in terms of provisioning node failure detection and membership services for CAN FD systems.

10:25
Paulo Barbosa (CISTER Research Centre, Portugal)
António Barros (CISTER Research Centre, Portugal)
Luis Miguel Pinho (CISTER Research Centre/ISEP, Portugal)
IoT for Energy-Efficiency: connecting a serious game with energy metering in the EnerGAware project
SPEAKER: unknown

ABSTRACT. Cyber-physical systems and the internet of things push for a multitude of devices and systems, needing to work together to provide the required services. However, the speed of development and the heterogeneity of devices introduces considerable challenges in the development of such systems. This communication describes a solution being implemented in the setting of a serious game scenario, connected to real homes energy consumption. The solution provides a publish-subscribe middleware which is able to seamlessly connect all the components of the system.

09:30-10:30 Session 6B: Gestão de Dados e Conhecimento - GDC
Chair:
Helena Galhardas (Instituto Superior Técnico, Portugal)
Location: Sala 2
09:30
João L. M. Pereira (INESC-ID and Instituto Superior Técnico, Universidade de Lisboa, Portugal)
Helena Galhardas (INESC-ID and Instituto Superior Técnico, Universidade de Lisboa, Portugal)
Approximate Duplicate Elimination using State-Of-The-Art Tools: a Comparison
SPEAKER: unknown

ABSTRACT. Successfully analyzing data is a challenge for the scientific community and companies since raw data usually contains several data quality problems that influence data-based decisions. The presence of records that refer to the same entity in the real world is one of the most common data quality problems that greatly impact data analytics. In order to guarantee a unique and proper representation record for each entity, an Approximate Duplicate Elimination procedure has to be performed. It identifies records that refer to the same entity and then generates a unique record representation for each entity. Distinct commercial and research tools have been developed to carry out Approximate Duplicate Elimination. However, there is no systematic comparison of tools regarding their support for Approximate Duplicate Elimination.

This paper aims at providing a comparison of three different types of tools that can perform the Approximate Duplicate Elimination task, namely Pentaho Data Integration, CLEENEX, and OpenRefine. For this purpose, in order to support an experimental evaluation, we generated a synthetic dataset with errors, we specified the same Approximate Duplicate Elimination task for each tool, and then we evaluated these three tools using different criteria. With this comparison, we expect that users are able to choose the most appropriate tool or a combination of tools that better suit a specific Approximate Duplicate Elimination task.

09:50
André Pires (INESC TEC, Portugal)
José Devezas (DEI, Faculdade de Engenharia, Universidade do Porto, Portugal)
Sérgio Nunes (INESC TEC and FEUP, U.Porto, Portugal)
Benchmarking Named Entity Recognition Tools for Portuguese
SPEAKER: unknown

ABSTRACT. There has been much work in the Information Extraction field, particularly in the Named Entity Recognition task. However, for the Portuguese language, the implementations still perform below the results for other languages, as shown by the HAREM conferences. The goal of this work is to assess the current performance of well established tools, namely Stanford CoreNLP, OpenNLP, spaCy and NLTK, against a Portuguese dataset, specifically the HAREM golden collection. These tools were used with an out-of-the-box approach, meaning without any tuning. The results show that the referred tools can match the results of Portuguese state-of-the-art tools, presented in the HAREM conferences, specifically, Stanford CoreNLP with an F-measure of 56.10%, OpenNLP with 53.63%, spaCy with 46.81% and NLTK with 30.97%. Furthermore, a hyperparameter study was performed, improving over the default configurations, for about 2% in each tool, and almost 35% in NLTK's MaxEnt classifier.

10:10
Jorge Goulart (INESC-ID and IST, Portugal)
Pedro Silva (INESC-ID and IST, Portugal)
Alberto Vale (IPFN and IST, Portugal)
João Dias Pereira (INESC-ID and IST, Portugal)
Helena Galhardas (INESC-ID and IST, Portugal)
Towards an Information Management System for a Research Lab
SPEAKER: unknown

ABSTRACT. An Information Management System (IMS) is a software platform that enables the management of a vast array of information about products, customers, employees, suppliers, projects, production, assets and finances of an organization. Scientific research labs and education centers also benefit from the use of an IMS since they need to manage information about their employees, projects and assets. IPFN is a research unit associated to IST, and is an example of a scientific research lab. The current IMS of IPFN uses outdated technologies and cannot satisfy the new requirements that emerged through the time. Therefore, IPFN needs a new IMS in order to replace the current one.

This paper presents the design, implementation and validation of the new IMS for IPFN. The system is modular in nature, and each module is independent from others. This allows the system to be easily extended in the future with new modules. Additionally, the system was designed to be integrated with other external software systems belonging to IPFN and IST. We performed a set of performance and usability tests to ensure that the system satisfied the IPFN requirements. The new system was placed in production on March 2017. We report the results of the tests performed as well as the main results obtained after 3 months of use in a real environment.

10:15
Leandro Ricardo (Instituto de Telecomunicações, Aveiro, Portugal, Portugal)
Susana Sargento (Instituto de Telecomunicações, Aveiro, Portugal, Portugal)
Ilídio Oliveira (Instituto de Engenharia Electrónica e Telemática de Aveiro, Aveiro, Portugal, Portugal)
Decision Support System for City Public Transportation
SPEAKER: unknown

ABSTRACT. Porto, the second largest city in Portugal, is a Smart City. It deploys a world-level pilot project implementing a Vehicular Ad-hocNetwork connecting buses, garbage trucks and taxis. This mesh network is also capable of delivering free wireless internet access to bus passengers, enhancing their daily commuting experience. These wandering nodes of the vehicular network generate spatiotemporal information regarding their movement, making it possible to better understand how the bus network generally behaves and deliver new solutions for the main stakeholders of this use-case: the bus fleet managers and the bus passengers. The purpose of this study is to develop an integrated solution which is able to present time estimations and predictions to bus passengers and deliver a framework for helping bus fleet managers on understanding the quality of the bus network service.

10:30-11:00Coffee Break
10:30-11:00 Session 7: POSTERS (Continuação)
Chair:
Miguel Pupo Correia (INESC-ID, Instituto Superior Técnico, Universidade de Lisboa, Portugal)
11:00-13:00 Session 8A: Ciência e Engenharia de Software - SOFT-PT
Chair:
João Costa Seco (Universidade Nova de Lisboa, Portugal)
Location: Salão Nobre
11:00
José Melo (FEUP, Portugal)
Gabriel David (INESC TEC, Faculdade de Engenharia, Universidade do Porto, Portugal)
Anonymization of clinical data
SPEAKER: unknown

ABSTRACT. Clinical research and public health studies require extensive and detailed access to the vast amount of health records being collected in most health related services. By studying clinical data, researchers are able to produce statistics, identify relations and learn trends in the health sector. These studies and data analysis are very important as they can bring great benefits and knowledge in healthcare. However, maintaining individual privacy is crucial for ethical and legal reasons. To balance the need for rigorous data and the requirement of privacy protec- tion, a study with a prototype on data anonymization is done and some models and algorithms are discussed. This study allows to propose and develop a prac- tical way to efficiently anonymize clinical data. With this solution, the user can quickly and easily anonymize a given MongoDB dataset,through the provision of a set of configurations. The anonymization is done resorting to well known models and algorithms to protect privacy, associated with specific clinical crite- ria, restrictions and hierarchies. At the end of the anonymization, an anonymized version of the subset is obtained that meets the selected privacy model, balancing enough privacy versus keeping research value. To solve some limitations imposed by the implemented prototype, some op- timized versions are also presented. These optimizations allowed to improve per- formance and handle bigger datasets. All versions are compared in terms of performance and characteristics is done to better understand and decide which is the best version to a specific case. Quantified metrics of information loss are provided.

11:20
Fábio Ferreira (Large-Scale Informatics Systems Laboratory, Portugal)
Telmo Santos (Large-Scale Informatics Systems Laboratory, Portugal)
Francisco Martins (LaSiGE and Dept. of Informatics, Faculty of Sciences, University of Lisbon, Portugal)
Antónia Lopes (LaSiGE and Dept. of Informatics, Faculty of Sciences, University of Lisbon, Portugal)
Vasco T. Vasconcelos (LaSiGE and Dept. of Informatics, Faculty of Sciences, University of Lisbon, Portugal)
Especificação de Interfaces Aplicacionais REST
SPEAKER: unknown

ABSTRACT. A programação de serviços web que oferecem interfaces aplicacionais REST é atualmente muito popular, sendo que o desenvolvimento eficaz de aplicações cli- entes destes serviços exige que as suas interfaces estejam bem documentadas. No entanto, e apesar de iniciativas importantes como a Open API Specification, o su- porte à descrição destas interfaces é atualmente ainda bastante limitado, focado essencialmente nos aspetos sintáticos. Neste artigo apresenta-se a linguagem HEADREST que permite ultrapassar estas limitações e descrever também os aspetos semânticos destes interfaces num estilo reminescente dos triplos de Hoare e recor- rendo a tipos refinados. A linguagem é apresentada através de pequenos exemplos extraído de um dos casos de estudo desenvolvidos para a avaliar. Discute-se ainda a forma de validar a boa formação das especificações HEADREST e uma técnica para a geração de testes para testar a conformidade de uma API REST relativa- mente à sua descrição.

11:40
Alcides Fonseca (Universidade de Lisboa, Portugal)
Bruno Cabral (University of Coimbra, Portugal)
Fork-Work-Join: a model for executing Fork-Join programs on the GPU
SPEAKER: unknown

ABSTRACT. GPUs have become a popular tool for parallel programming. When compared with multicore architectures, GPUs can offer higher throughputs. Typically, only data-parallel programs can be efficiently ported to GPUs. On the other hand, task-parallel programs often produce many divergent paths of execution, which raise concurrency problems resulting in an inefficient scheduling of GPU resources.

One of the most popular approaches to express task-parallel programs is the Fork-Join model, which fits perfectly on divide-and-conquer algorithms and typically makes an extensive use of recursive calls, which can be served simultaneously by multiple processors. Since generic GPU programming languages, such as CUDA and OpenCL do not support recursion, it is not easy to express fork-join programs on the GPU. Ideally, one should be able to express programs independently of the target platform and still make an efficient use of the available computing resources.

We introduce a new approach, which enables an efficient usage of the fork-join model in programs targeting both CPU and GPU architectures. The programming model combines three operations, which are supported by different backends in each platform: Fork, Join and Work. Fork and Join are lightweight operations that split calls or merge their results, while Work performs the heavyweight computations and enables recursion.

We have implemented two popular benchmarks, Fibonacci and Integral, showing that this approach can be used to express those types of programs. Those programs were executed on 2 different systems and the GPU version outperformed the multicore version of Fork-Join. We were able to achieve 2 times the speedup of the ForkJoin CPU version with 24 workers.

12:00
Miguel Pasadinhas (INESC-ID, Portugal)
Daniel Porto (INESC-ID, Portugal)
Antónia Lopes (LASIGE, Portugal)
Luis Rodrigues (INESC-ID, Portugal)
Adaptação Guiada por Políticas de Sistemas Tolerantes a Faltas Bizatinas
SPEAKER: unknown

ABSTRACT. Ataques maliciosos, erros de software ou até mesmo enganos de operadores podem causar desvios arbitrários do comportamento esperado dos sistemas. A tolerância a faltas bizantinas (BFT) engloba um conjunto de técnicas que tornam os sistemas robustos na presença de faltas arbitrárias. Vários protocolos BFT foram propostos, cada um otimizado para diferentes condições operacionais e geralmente com fraco desempenho fora dessas condições. Para colmatar esse problema, propomos o Policaby, um gestor de adaptação capaz de executar políticas de adaptação que guiam o sistema num caminho de conformidade com os seus objetivos. Estas políticas são descritas numa linguagem de alto nível, com a expressividade necessária para capturar as adaptações desejáveis neste tipo de sistemas.

12:20
Tiago B. Fernandes (University of Minho, Portugal)
António Nestor Ribeiro (University of Minho, Portugal)
David V. Nunes (OutSystems, Portugal)
Hugo R. Lourenço (OutSystems, Portugal)
Luiz C. Santos (OutSystems, Portugal)
Support for Automatic Refactoring of Business Logic

ABSTRACT. Software’s structure profoundly affects its development and maintenance costs. Poor software’s structure may lead to well-known design flaws, such as large modules or long methods. A possible approach to reduce a module’s complexity is the Extract Method refactoring technique. This technique allows the decomposition of a large and complex method into smaller and simpler ones, while reducing the original method’s size and improving its readability and comprehension.

The OutSystems platform is a low code platform that allows the development of web and mobile applications that rely on a set of visual Domain-Specific Languages (DSLs). Even low code languages when improperly used can lead to software that has maintenance issues like long methods.

Thus, the purpose of this paper is to present the research and development done to provide the OutSystems platform with a tool that automatically suggests Extract Method refactoring opportunities. The research combines program slicing techniques with code complexity metrics to calculate the best refactoring opportunities that preserve programs’ functionality.

The proposed approach was tested on typical OutSystems apps and was shown to be able to reduce the overall applications’ complexity.

12:40
André Barbosa (Instituto Superior Técnico, Portugal)
Ana Costa (Instituto Superior Técnico, Portugal)
Francisco Bom (Instituto Superior Técnico, Portugal)
José Figueira (Instituto Superior Técnico, Portugal)
José Borbinha (Instituto Superior Técnico, Portugal)
DecSpace: A Multi-Criteria Decision Analysis Framework
SPEAKER: unknown

ABSTRACT. This paper describes a web-based application to explore Multi-Criteria Decision Aiding (MCDA) methods. MCDA is an important branch of Operational Research, which deals with decision aiding problems involving multiple criteria. It aims at helping decision-makers during the decision aiding process. MCDA offers a large range of approaches, methods, and techniques to deal with the decision situations at hand. Due to the considerable complexity of the interactions or amount of computation required by some MCDA methods, applying such methods without software support is hardly imaginable. Therefore, Decision Support Systems (DSS) are required for designing and executing MCDA methods. The increasing interest of researchers in this domain is leading to the growth of the number of MCDA methods and, therefore, to the number of those DSS systems. Most of those software solutions are developed to fit the specific needs of the interested researchers, without considering possible uses outside of their narrow fields, therefore they usually support only one or two MCDA methods. DecSpace is a web-based DSS that aims to provide an easy-to-use and intuitive approach of using any kind of MCDA method, independently of its complexity. The user-friendliness and accessibility are some of the most important design concerns of this system, since it not only intends to make it possible to use the methods, but also to facilitate their understanding and learning by non-expert users. DecSpace also allows the creation of workflows that may use multiple MCDA methods, so that the outputs of a certain method can be used as input to another one and, therefore, making it possible to design and experiment sophisticated decision-aiding processes. DecSpace also implements the diviz open architecture, making it possible to use an already existing extensive collection of MCDA methods and to easily add new ones.

12:45
Paweł Krysiak (FCT, Portugal)
Hugo Lourenço (OutSystems, Portugal)
A Data Validation DSL for the OutSystems language (Comunication)
SPEAKER: unknown

ABSTRACT. The correctness of applications' code is always based on strong assumptions about the quality of their data, that is either stored in databases or derived from user input. The data lifecycle always starts on validation conditions performed at user input. Validation code is usually crafted and replicated into web forms, web services code, and embedded deep into the logic of the application. Its task is to make sure that no inconsistent information enters the system. The way that web and mobile applications are structured forces us to validate data on both the client and server parts of the application. Ensuring correctness on the dierent layers of the system implies the duplication of business rules, which leads to more complex software architecture and more error prone code. Forms are one example of an input data source to the system, and will be the main focus of this paper. We start by presenting a textual language inspired by the OutSystems visual DSL, that allows us to create web applications which have client-side and server-side code. The language is extended to allow the developer to, using a declarative approach, write validation rules that can be applied to data types or input elds in a form. The compiler ensures that these validation rules are enforced at all layers of the application, namely in both the client and server side.

12:50
Ana C. R Paiva (FEUP / INESC TEC, Portugal)
Nuno H. Flores (FEUP / INESC TEC, Portugal)
João P. Faria (FEUP / INESC TEC, Portugal)
José Manuel Granate Marques (Link, Portugal)
Automatic Business Process Testing
SPEAKER: unknown

ABSTRACT. End-to-end business process automatic validation can be a challenging task, yet an important way to check that business rules are working properly and that problems are detected and corrected as soon as possible. This paper presents ETAP-PRO, an end-to-end test automation platform for business processes that aims to overcome some challenges in validating business processes. Its goals, architecture and functionality are briefly described.

11:00-13:00 Session 8B: Computação Móvel e Ubíqua - CMU
Chair:
Nuno Cruz (Instituto Politécnico de Lisboa, Portugal)
Location: Sala 2
11:00
Pedro Sanches (Faculdade de Ciências e Tecnologia - Universidade Nova de Lisboa, Portugal)
Hervé Paulino (Faculdade de Ciências e Tecnologia - Universidade Nova de Lisboa, Portugal)
Filipe Cerqueira (Faculdade de Ciências e Tecnologia - Universidade Nova de Lisboa, Portugal)
João Silva (Faculdade de Ciências e Tecnologia - Universidade Nova de Lisboa, Portugal)
António Teófilo (ISEL - Instituto Superior de Engenharia de Lisboa, Instituto Politécnico de Lisboa, Portugal)
Computação distribuída em redes formadas por dispositivos móveis
SPEAKER: unknown

ABSTRACT. Nos últimos anos temos visto um aumento significativo, tanto no número de dispositivos móveis e nas suas capacidades computacionais, de armazenamento e de comunicação, como também no número de aplicações que necessitam de cada vez mais recursos computacionais e de armazenamento. Atualmente, de forma a lidar com esta necessidade crescente de recursos, estas aplicações fazem uso de serviços Cloud, oferecidos por infraestruturas Cloud. No entanto, com a utilização destes serviços, advêm alguns problemas, como o elevado valor de latência experienciado quando um dispositivo móvel estabelece uma conexão com uma destas infraestruturas, e o considerável uso de energia e largura de banda. Também a necessidade de manter uma conexão à Internet pode ser um obstáculo em situações onde infraestruturas de conectividade se encontram congestionadas ou simplesmente não existem. Considerando o que foi abordado acima, para algumas aplicações começa a fazer sentido efetuar parte ou toda a computação localmente nos próprios dispositivos móveis. Neste artigo apresentamos Oregano, um sistema de computação distribuída centrado nos dados, para dispositivos móveis. Este sistema é uma framework, capaz de processar conjuntos ou fluxos de dados, que são gerados por redes compostas por dispositivos móveis, sem necessitar de uma conexão à Internet. Contrariamente ao estado de arte atual, onde a computação e os dados são enviados para dispositivos móveis, o nosso sistema tem como objetivo mover a computação para onde os dados se encontram, reduzindo significativamente a quantidade de dados trocados.

11:20
Nuno Sousa (Instituto Superior Técnico, Portugal)
João Nuno Silva (INESC-ID, Portugal)
UBILocus - Framework for location aware application development
SPEAKER: unknown

ABSTRACT. Nowadays the number of location-aware applications and services is increasing. The development of these applications relies on the Operating System provided API's or in libraries provided by specific positioning providers. The use of multiple positioning providers in the same application (for instance GPS and Beacons based indoor positioning) requires the use and integration at the application level of various (and sometimes incompatible) functions and data types. Furthermore, the use of complementary services requires the integration with respect to the coordinate system to use.

In this paper, we propose a middleware that allows the integration and abstraction of multiple positioning providers. We defined the library interface, the application data types, describe its implementation and on usage application. Besides being a positioning provider (using multiple technologies), UBILocus also integrates a route generation module usable by the mobile application.

UBILocus was implemented using the Xamarin Platform, allows the use of two positioning providers (GPS and Beacons), provides two routing subsystems and seamlessly presents in the application private maps. The implemented UBILocus was tested in an Android application.

11:40
Pedro Alcobia (IST, Portugal)
João Barreto (INESC-ID, Portugal)
sistema de guarda de bicicletas em meio urbano
SPEAKER: unknown

ABSTRACT. Hoje em dia é complicado estacionar a bicicleta na cidade sem nos preocuparmos com a sua segurança. As soluções para aumentar a segurança nos estacionamentos visam o uso de estruturas complexas de arrumo de bicicletas, aumentando assim o custo associado a estes estacionamentos. Em Portugal, os estacionamentos para bicicletas, em geral, são simples e fornecem pouca segurança. Nesta tese propomos um sistema de guarda de bicicletas com vista a aumentar a segurança das mesmas quando estacionadas. Tal é conseguido através da monitorização das bicicletas no estacionamento. Esta monitorização de bicicletas será feita através de um sistema informático que compreende: a presença de pequenos beacons Bluetooth nas bicicletas com o intuito de identificar e comunicar a sua presença no estacionamento; um dispositivo fixo no estacionamento que monitoriza as bicicletas ao seu alcance enviando esta informação para um servidor central; um servidor central que analisa entradas e saídas de bicicletas, comunicando ao dono caso a bicicleta tenha saído indevidamente do estacionamento; uma interface gráfica que receberá os alertas de saída indevida de bicicleta comunicados pelo servidor central e irá notificar devidamente o dono. Esta abordagem permite uma segurança mais ativa no estacionamento de bicicletas com menor impacto nas infraestruturas necessárias para tal. Em caso de Roubo, a nossa solução, por ter como base uma solução de localização e monitorização de bicicletas, pode também permitir o encontro de bicicletas roubadas. No desenvolvimento desta solução são levantadas algumas questões relevantes ao nível do consumo energético do beacon, que irá identificar a bicicleta, e da segurança\privacidade do utilizador ao usar o nosso sistema. Ambas estas questões prendem-se com o beacon que será usado para identificar a bicicleta. A solução proposta nesta tese foca-se na privacidade e disponibilidade do sistema em si, permitindo ao utilizador monitorizar a sua bicicleta quando estacionada.

12:00
Diogo Lima (Lasige, Universidade de Lisboa, Portugal)
Hugo Miranda (Universidade de Lisboa - Faculdade de Ciências - LaSIGE, Portugal)
Francois Taiani (Université de Rennes 1 - ESIR / IRISA, France)
Can Graphs Solve the Geo-aware State Deployment Problem?
SPEAKER: unknown

ABSTRACT. A challenge for large scale distributed mobile applications, such as augmented reality games and social networks, is the management of application state. Because the infrastructure is concentrated in datacenters geographically distant from the end users, state flowing between the large number of users and the supporting infrastructure suffers from latency and network congestion degrading user quality of experience.

To mitigate this problem, Fog Computing proposes a physical approximation of the servers to the end users, avoiding the use of the Internet backbone. However, gains depend of the system capability to correctly deploy each state component at its most convenient location. Geo-aware state deployment is challenging as it requires a continuous observation and interpretation of the state utilization patterns, expected to evolve with time as new trends and user behaviour emerge. The paper proposes a new object graph-based approach for the geo-aware state deployment problem and compares its performance with approaches based on linear algorithms.

12:20
Luis Afonso (Universidade de Aveiro, Portugal)
Mariana Souza (University of São Paulo, Brazil)
Paulo Dias (Universidade de Aveiro, Portugal)
Joaquim Madeira (Universidade de Aveiro, Portugal)
Margarida Marques (Universidade de Aveiro, Portugal)
Lucia Pombo (Universidade de Aveiro, Portugal)
EduPARK app: a evolução de uma aplicação móvel para aprendizagem em contexto outdoor
SPEAKER: unknown

ABSTRACT. As tecnologias móveis têm sido cada vez mais utilizadas no cotidiano das populações, porém ainda são pouco exploradas para fins educacionais, quer em contexto de sala de aula quer em contexto outdoor. Visando promover a aprendizagem outdoor, o projeto EduPARK está a desenvolver uma aplicação móvel que utiliza Realidade Aumentada (RA) e estratégias de gamificação dirigida a alunos, professores e público em geral. Desta forma, é possível explorar um parque citadino com recurso a conteúdos adicionais ao real, como textos, áudios, vídeos e modelos em 3D. A aplicação permite interagir, durante uma visita ao parque, com informação de caráter multidisciplinar, integrando as orientações curriculares, que ajuda a aprender sobre a cidade e o parque, eventos históricos, botânica local, entre outros. O objetivo deste artigo é apresentar os primeiros ciclos de desenvolvimento da aplicação móvel EduPARK, reportando as alterações que têm vindo a ser introduzidas com base no feedback dos utilizadores. Durante as sessões de teste com os utilizadores foi recolhida informação relativa ao desenho, usabilidade, funcionalidade e conteúdo, essencial para a evolução do protótipo. O projeto EduPARK enquadra-se numa metodologia de design-based research que prevê vários ciclos de melhoria e refinamento da aplicação. A recolha de dados foi efetuada com recurso a focus group e questionário administrado aos utilizadores após a experimentação da aplicação. Os resultados indicam que a conjugação das tecnologias móveis com Realidade Aumentada (RA) e estratégias de gamificação, apresenta novos desafios e oportunidades, em particular, na área educacional, potenciando a aprendizagem e estabelecendo ligação com o conteúdo curricular.

12:25
Filipe Cerqueira (Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa, Portugal)
Hervé Paulino (Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa, Portugal)
João A. Silva (Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa, Portugal)
João M. Lourenço (Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa, Portugal)
Um Sistema Publicador/Subscritor com Persistência de Dados para Redes de Dispositivos Móveis
SPEAKER: unknown

ABSTRACT. Os dispositivos móveis são atualmente uma das principais fontes de geração de conteúdos pessoais. A portabilidade e a crescente capacidade destes dispositivos permitem que os mesmos sejam utilizados em variadas atividades, lúdicas e não só. Tal tendência tem sido acompanhada pelo progressivo interesse em partilhar (em tempo real) os conteúdos gerados. No entanto, a comum utilização de serviços centralizados para suportar a partilha de dados entre utilizadores requer sistemas de servidores com grande capacidade de resposta e coloca muita carga na rede, reduzindo a sua largura de banda com dados que (muitas vezes) podem ser partilhados na periferia. Neste artigo apresentamos o Thyme, um sistema publicador/subscritor com persistência de dados para redes de dispositivos móveis, que fornece um mecanismo de disseminação e armazenamento de dados adaptado para este tipo de redes, sem requerer acesso à Internet. No Thyme, o tempo é uma dimensão de primeira ordem, associando um intervalo de tempo de vida a cada subscrição, em que o início e fim do intervalo podem ser referentes ao futuro, presente ou passado. Como caso de uso apresentamos uma aplicação Android para partilha de fotografias. Os resultados mostram tempos de resposta que garantem a interatividade com o utilizador, e um baixo consumo de energia, permitindo que a aplicação seja utilizada sem gastos excessivos de bateria.

12:30
Miguel Machado (Universidade de Aveiro, Portugal)
António Borges (Universidade de Aveiro, Portugal)
António Oliveira (WithUS Lda., Portugal)
A Monitoring System Based on Fog Computing
SPEAKER: unknown

ABSTRACT. In a pervasive computing context, the common approach assumes a cloud computing paradigm as the central provider of management and processing power to a sensor network. This design implies, however, a potentially enormous amount of data exchanged directly between the cloud and the sensor nodes. Thus, impairing communication and requiring extra levels of software complexity to make it efficient. Fog Computing has emerged as a new design paradigm to these types of systems, aiming to deliver a homogeneous service while removing most of the drawbacks previously described. Through this approach, an extra layer of computing power is placed between the cloud and the sensor network, acting both as a mediator whose central task is to manage, monitor and collect data from geographically-located groups of sensor nodes and as a communication hub to the cloud with which data is exchanged in a compact and minimalist fashion. This paper describes a system comprised of nodes with autonomous behavior, able to organize themselves into smaller groups within the system. The inherited mechanisms and behavior ultimately provide fault tolerance to the overall structure, maintaining the operational status quo. The solution is supported and was validated by a proof of concept where actual computing devices were used to partition a set of sensor nodes into local groups and to collect data from them, delivering an ubiquitous service while there is, at least, one computing device operating. The experiment assumed a distributed mains power measurement at different sockets within a building, where the sensor nodes were implemented by smart plugs and the fog layer was implemented by smart gateways running a Java virtual machine under Linux.

12:35
Bruno Areias (Instituto de Telecomunicações, Portugal)
Nuno Humberto (Instituto de Telecomunicações, Portugal)
Lucas Guardalben (Instituto de Telecomunicações, Portugal)
Susana Sargento (Instituto de Telecomunicações, Portugal)
Automated Flying Drones Platform for Sensing Smart Spaces
SPEAKER: unknown

ABSTRACT. }For a long time Unmanned Aerial Vehicles (from now on designated as \textit{drones}) have been used within military forces (e.g., aerial surveillance and high risk bombing mission). Due to the evolution of technology, drones are becoming cheaper and accessible to the general population, used in various tasks (e.g., aerial photography, package/mail delivery, building inspections, sports...). Currently, automation is already possible (e.g., GPS way-point, go home and path flight), but requires direct drone connectivity and an extensive knowledge over technical details of the drone, such as its construction, operation and configuration of the flight parameters. This paper proposes a platform to offer drones as a service for data collection in Smart Spaces, abstracting the end-user of the technical aspects and limitations of the drone.

12:40
Daniela Simões (University of Aveiro/Instituto de Telecomunicações, Portugal)
Lucas Guardalben (Instituto de Telecomunicações, Portugal)
Susana Sargento (University of Aveiro/Instituto de Telecomunicações, Portugal)
A Drone-Assisted Service for Emergency Scenarios
SPEAKER: unknown

ABSTRACT. Unmanned Aerial Vehicles (UAVs), as popularly known by Drones, are already part of our everyday life, mainly due to their attractive cost and popularity. Beyond military applications, where Drones have broad prominence, numerous applications have been proposed for many different areas (e.g., search/rescue missions, construction/inspection, product/food delivery, surveillance). At this stage, Drones are enabling a new set of applications and services for civilians and security forces (e.g., firemen, policemen) with exclusive aerial perspectives, enhancing search/rescue/surveillance missions over emergency scenarios.

This paper presents an overview of the proposed drone-assisted emergency service, which is a valuable support for civilians and special security forces in case of emergency situations.

12:45
João Ribeiro (Instituto de Telecomunicações, Portugal)
André Zúquete (Instituto de Telecomunicações, Portugal)
Susana Sargento (Instituto de Telecomunicações, Portugal)
Survey of public transport routes using Wi-Fi
SPEAKER: unknown

ABSTRACT. An important aspect in improving public transport efficiency is collecting information regarding travelers’ routes, usually represented as an Origin Destination (OD) matrix. Most public transportation systems implement fare ollection systems that can provide the accurate origins of travelers’ routes but not accurate destinations. In this paper we look at Wi-Fi, more specifically 802.11 data-link layer, as a candidate to provide OD matrix estimations.

We present a system and an algorithm capable of collecting information regarding Wi-Fi capable devices inside the bus complemented with positioning and time. A system is also presented to implement this concept using minimal requirements.

An implementation of this system was deployed in a public bus to collect data for several weeks. This resulted on over 15000 traveler routes collected in 9 different days. This data was contextualized and mapped to an OD system in order to demonstrate how it can be used to generate OD matrix estimations.

13:00-14:00 Session : Almoço
Location: Zona de Restauração
14:00-15:00 Session 9: Orador Convidado 2
Chair:
António Teixeira (University of Aveiro, Portugal)
Location: Auditório Amorim
14:00
Liliana Ferreira (FhP, Portugal)
Information in healthcare: a healthier future

ABSTRACT. Understanding the users, their activities, context and behavior and providing tailored recommendations to patients and health professionals are some of the potential applications of automatic tools that extract information from structured and unstructured data sources and perform accurate automatic mapping of free data onto a structured knowledge representation.

This talk provides a short walk through the opportunities and challenges of the application of information and communications technologies to the delivery of safe, effective, timely, patient-centered, and efficient health care.

 

15:00-17:05 Session 10A: Segurança de Sistemas de Computadores e Comunicações - SSCC
Chair:
André Zúquete (IEETA, University of Aveiro, Portugal)
Location: Sala 2
15:00
Pedro Do Vale (INESC-ID, IST, Universidade de Lisboa / Multicert, Portugal)
Carlos Cardoso (Multicert, Portugal)
Renato Portela (Multicert, Portugal)
Ricardo Chaves (INESC-ID, IST, Universidade de Lisboa, Portugal)
RSIGN: Secure Remote Qualified Signature System
SPEAKER: unknown

ABSTRACT. Qualified signatures (as defined in the EU directive 1999/93/EC) are equivalent to hand written signatures. To assure this, they need to be created in a secure environment in order to grant them a strong legal value. The most common approach to accomplish this is with the use of secure devices, such as smart cards, to store and use the associated private keys. To avoid users having to carry such devices, this paper presents a solution providing remote qualified signatures on mobile devices. This is accomplished by having a remote service, able to create qualified signatures for remotely connected users. The implemented solution is able to provide a secure connection between the user device and the qualified zone in the server infrastructure, supporting multiple users each with unique private keys managed by a Hardware Security Module. This solution allows for qualified signatures creation in a scalable and secure manner, providing a commercially and technically attractive solution.

15:20
Rui Ribeiro (Universidade de Aveiro, Portugal)
Miguel Azevedo (Universidade de Aveiro, Portugal)
André Zúquete (Universidade de Aveiro, Portugal)
NFC-based, off-line door lock access control mechanism
SPEAKER: unknown

ABSTRACT. This paper proposes an NFC-based, off-line door access control mechanism. The credentials to access the door services, such as opening its lock, are first obtained by users from a central authentication server. Those credentials are time bounded, i.e., they can only be used on a specific time frame. In our case, we used daily credentials, but other time intervals could have been used. Door locks keep a record of all successful opening requests, which can be collected using another door service, also available through its NFC interface. The access credentials are fetched and used, within a challenge-response authentication protocol, by an application running on an NFC-enabled device, such as a smartphone. For a proof of concept we implemented a prototype of our system with an Arduino with an NFC shield, for controlling a door lock, an Android Nexus 7 mobile device and a Web server for providing daily access credentials.

15:40
Fernando Mário M. Marques (Instituto Superior Técnico, Portugal)
Ana Almeida Matos (Instituto Superior Técnico, Portugal)
Jan Cederquist (Instituto Superior Técnico, Portugal)
Integrating paper-based voting and Belenios -- a hybrid voting protocol for an academic organization

ABSTRACT. We consider the problem of introducing an electronic voting system in the context of an academic organization. We work under the assumption that such a context fits a profile where integrity of an election is expected to be fully verifiable, privacy must be ensured by the system, and coercion is not of primary concern. Additionally, we assume that a classic paper-voting is already in place, and while the voting community may be expected to be literate in handling web-based applications and in understanding the concept of verifiability, it may exhibit some cultural resistance against shifting to a new voting practice. We present an integration of a classical paper ballot system with the Belenios e-voting protocol, an on-line voting system which guarantees vote privacy and verifiability. To this end, we propose a specification of a hybrid version of the Belenios protocol which integrates the classical paper ballots. We then present and prove, up to assumptions of security of the two baseline protocols, the relevant security properties of the resulting protocol. Finally, we show that the proposed protocol is well suited to be adopted by an academic organization by presenting an architecture for the proposed solution that fits the current requirements of an existing university.

16:00
Ivo Vacas (LaSIGE, Faculdade de Ciências da Universidade de Lisboa, Portugal)
Ibéria Medeiros (LaSIGE, Faculdade de Ciências da Universidade de Lisboa, Portugal)
Geração Automática de Conhecimento para SDI extraído de OSINTs
SPEAKER: unknown

ABSTRACT. O cibercrime a organizações tem sido uma constante nos dias de hoje. As organizações para se protegerem destes ataques utilizam mecanismos de defesa, tais como sistemas detetores de intrusões (SDI), no entanto, a sua eficácia na deteção destes ataques depende do conhecimento que estes contêm sobre as ameaças e da forma como as deteta. O contínuo aparecimento de novos e sofisticados ataques torna obrigatório que os SDI sejam atualizados constantemente com conhecimento sobre novas ameaças. Este conhecimento pode ser obtido de diversas fontes de inteligência -- Open Source Intelligence (OSINT) -- públicas, as quais se encontram acessíveis em diversos locais na internet. Este artigo apresenta uma solução para melhorar uma arquitetura de deteção de intrusões em SDI. A solução propõe um gerador de regras e blacklists para o SDI, com base em informação OSINT recolhida por uma plataforma de threat inteligence bem como a sua integração no SDI de forma automática. Foi realizada uma avaliação experimental da solução em ambiente real, usando 44 fontes de OSINT recolhidas pela plataforma threat inteligence IntelMQ e o SDI Snort. A arquitetura proposta permitiu detetar ameaças de diversas categorias.

16:20
Raimundo Chipongue (Faculdade de Ciências da Universidade de Lisboa, Portugal)
Hugo Miranda (Faculdade de Ciências da Universidade de Lisboa, Portugal)
António Broega (Faculdade de Ciências da Universidade de Lisboa, Portugal)
Indicadores de Segurança em Plataformas de Monitorização
SPEAKER: unknown

ABSTRACT. A separação entre administradores de sistemas e redes (TI) e especialistas de segurança afeta a capacidade e tempo de resposta a incidentes de muitas organizações. Esta separação é também observável na limitada oferta de módulos relacionados com segurança nos repositórios públicos de plataformas de monitorização. No entanto, a integração da visão de administração de IT com a da equipa de segurança, por exemplo nas consolas das plataformas de monitorização, pode contribuir para um diagnóstico mais eficaz e para a deteção precoce e resolução de incidentes.

Neste artigo discutimos a contribuição das ferramentas de monitorização no colmatar desta separação. Para tal, foi desenvolvido um conjunto de módulos para a plataforma Nagios, que procuram proativamente vulnerabilidades conhecidas e comportamentos anómalos, indicadores de potenciais problemas de segurança. A avaliação dos módulos em ambiente de produção e a sua aprovação pela comunidade de editores do repositório oficial de módulos do Nagios fazem-nos concluir que é possível usar a flexibilidade destas ferramentas para monitorização conjunta de sistemas, redes e eventos de segurança, facilitando e acelerando a sua deteção e resolução.

16:40
Diogo Monteiro (INESC-ID / Instituto Superior Técnico, Portugal)
Paulo Ferreira (INESC-ID / Instituto Superior Técnico, Portugal)
TrustedVote - Trusted Computing for Internet Voting
SPEAKER: unknown

ABSTRACT. The wide availability of mobile devices, such as smartphones, enabled mobility in our lifestyles. However, traditional voting systems require physical presence of the voter at a specific place and time, which is incompatible with the concept of mobility. The goal of this paper is to propose an Internet voting system, called TrustedVote, that allows voters to cast their vote anywhere. Moreover, Internet voting significantly raises the turnout rate, reduces administrative costs and tallying time. In order to tackle malware and other insecurities in the client mobile platform, the solution is based on smartphones with a Trusted Execution Environment (TEE). TrustedVote leverages the isolation properties of TEEs available in Android and iOS smartphones to perform the cryptographic steps of an Internet voting system, such as vote encryption and voter authentication.

16:45
Miguel Falé (LaSIGE, Faculdade de Ciências da Universidade de Lisboa, Portugal)
Ibéria Medeiros (LaSIGE, Faculdade de Ciências da Universidade de Lisboa, Portugal)
Nuno Neves (LaSIGE, Faculdade de Ciências da Universidade de Lisboa, Portugal)
Resolução de Dependências Circulares em Inclusão de Código em Análise Estática de Código
SPEAKER: unknown

ABSTRACT. Hoje em dia as aplicações web desempenham um papel fundamental no acesso a uma miríade de serviços. Construir aplicações seguras tornou-se assim num aspeto crucial, contudo, a sua obtenção depende tanto dos conhecimentos de quem as elabora, bem como da correta utilização das linguagens de programação. O uso incorreto de funcionalidades das linguagens origina inconsistências na execução do código, tais como dependências circulares criadas pela inclusão de código de um ficheiro noutro recursivamente. As ferramentas de análise estática de código são utilizadas para detetarem este tipo de erros nos programas. Estas percorrem o código das aplicações, estando assim sensíveis às dependências circulares que provocam deteções inválidas (tanto falsos positivos como falsos negativos). Este artigo apresenta o problema das dependências circulares e uma solução para a sua resolução através de análise estática de código. A solução foi implementada na ferramenta WAP, a qual identifica dependências circulares em aplicações web e realiza a análise sem quebrar a sua execução. Uma avaliação experimental foi realizada com outras ferramentas em aplicações web e plugins.

16:50
Miguel Guerra (INESC-ID, Instituto Superior Técnico, Portugal)
Miguel Correia (INESC-ID, Instituto Superior Técnico, Portugal)
Benjamin Taubmann (Universität Passau, Germany)
Hans Reiser (Universität Passau, Germany)
ITZ: An Introspection Library for ARM TrustZone
SPEAKER: unknown

ABSTRACT. In this paper we explore ARM TrustZone, an extension of the ARM architecture that allows software running in such processors to be split in two worlds: the secure world for the security subsystem, and the normal world for everything else. This work aims to provide a solution for analyzing the security status of the normal world, where a common OS (e.g., Linux) and applications are executed. For this purpose, we explore virtual machine introspection (VMI), i.e., we leverage the TrustZone architecture to perform introspection from the secure world (which runs typically just a slim OS and a few security services), and analyze the normal world searching for malicious behavior. Ultimately, we focus on the creation of a TrustZone compatible introspection library, that provides memory access and management functions.

16:55
Nuno Tavares (IST/ULisboa/INESC-ID, Portugal)
Joao Nuno Silva (Instituto Superior Técnico - UTL / INESC-ID Lisboa, Portugal)
Trusted RTK using a Blockchain Ledger
SPEAKER: unknown

ABSTRACT. The evolution of GPS and its ever increasing usage has led to many different and vast applications, many of them requiring some degree of precision not provided by the base GPS service. The many aug- mentation systems that exist today require some degree of investment or expensive fees since they rely on expensive equipment or setups. There is, therefore, a need for a low cost system that can be used by these ap- plications, providing the required degree of precision. The main objective of this project is to create a network of scalable and open-source base stations that evaluate each other with a reputation system based on a blockchain ledger for Real Time Kinematics (RTK) positioning.

15:00-17:00 Session 10B: Computação Paralela, Distribuída e de Larga Escala - CPDLA
Chairs:
Miguel Matos (Universidade de Lisboa, Portugal)
Paula Prata (Universidade da Beira Interior, Portugal)
Location: Salão Nobre
15:00
Gonçalo Cabrita (NOVA LINCS / DI-FCT-Univ. NOVA de Lisboa, Portugal)
Nuno Preguiça (NOVA LINCS / DI-FCT-Univ. NOVA de Lisboa, Portugal)
Replicação não uniforme de objetos replicados
SPEAKER: unknown

ABSTRACT. A replicação é uma técnica chave no desenho de sistemas distribuídos confiáveis e eficientes. À medida que a informação cresce, torna-se difícil ou até impossível guardar todos os dados em todas as réplicas. Uma solução comum a este problema passa por utilizar técnicas de replicação parcial, onde cada réplica mantém apenas uma parte dos dados. Por consequência, cada réplica consegue apenas responder a um subconjunto das queries suportadas. Neste artigo, introduzimos o conceito de replicação não uniforme, onde cada réplica guarda apenas parte dos dados consegue responder a todas as queries suportadas. Aplicamos este conceito à consistência eventual e aos tipos de dados livres de conflitos, propondo um tipo de dados genérico onde as réplicas sincronizam por troca de operações. As nossas avaliações mostram que a replicação não uniforme é mais eficiente que a replicação tradicional, conseguindo reduzir o espaço de armazenamento utilizado, a quantidade de dados transmitidos, e aumentando a escalabilidade do sistema.

15:20
Godinho Quissico (INESC-ID, Portugal)
João Pedro Barreto (INESC-ID/Universidade de Lisboa, Portugal)
Daniel Castro (INESC-ID, Portugal)
PART: Árvore ART em Memória Persistente
SPEAKER: unknown

ABSTRACT. The emerging non-volatile memory (NVM) technology is expected to revolutionize computing. The fact that they are byte addressable and have performance comparable to that of DRAM make these memories potential successors for both DRAM and disks.

Database systems would gain with the increase of physical memory capacity, lower cost failure-consistency. However, data structure do not attain durability automatically, hence, further research on NVM aware data structure is of utmost importance.

System failure may corrupt the data structure consistent state, as cached updates eventually hit memory in any order. Trivial solutions tackle this problem by constantly flushing updates into log, which may lead to repeated flushes of the same memory position.

In this paper, we propose to implement ART-tree on persistent memory based-system with consecutive memory space allocation, using append-only technique, with execution of flush commands in batch in background at the end of each memory page.

15:40
Bernardo Palma (INESC-ID, Portugal)
Daniel Porto (INESC-ID, Portugal)
Luis Rodrigues (INESC-ID, Portugal)
Monitorização de Sistemas Tolerantes a Faltas Bizantinas para Suportar Adaptação Dinâmica
SPEAKER: unknown

ABSTRACT. Este trabalho descreve um sistema de monitorização, responsável por recolher informação sobre a execução de um protocolo de tolerância a faltas Bizantinas. Esta informação é usada para alimentar um gestor de adaptação (ele próprio replicado), responsável por optimizar em tempo de execução o funcionamento do sistema gerido. Uma vez que se pretende que todo o sistema seja tolerante a faltas Bizantinas, o sistema de monitorização necessita de: i) recolher valores de vários sensores e aplicar funções de agregação tolerantes a faltas que permitam descartar valores errados; ii) concretizar mecanismos de consenso que garantam que todas as réplicas correctas do gestor de adaptação aplicam as políticas numa vista coerente. Este artigo descreve e avalia as opções de desenho seguidas no desenvolvimento do sistema de monitorização, com o objectivo de assegurar a sua extensibilidade e capacidade de escala.

16:00
Carlos Carvalho (INESC-ID, Portugal)
Daniel Porto (INESC-ID, Portugal)
Luis Rodrigues (INESC-ID, Portugal)
Alysson Bessani (LASIGE, Portugal)
Adaptação Dinâmica de Protocolos de Consenso Bizantino
SPEAKER: unknown

ABSTRACT. O problema de resolver o consenso distribuído na presença de faltas Bizantinas tem recebido particular atenção nas últimas décadas. Existem hoje diversos protocolos para este efeito, cada um optimizado para condições de execução particulares. Uma vez que na maioria dos casos os sistemas reais operam em condições dinâmicas, importa desenvolver mecanismos que permitam adaptar os protocolos em tempo de execução ou substituir um protocolo por outro mais adequado às condições correntes. O problema da adaptação dinâmica de protocolos de consenso não é novo, mas a literatura é escassa para o caso Bizantino e não existem trabalhos que permitam comparar as soluções existentes. Este trabalho tem dois objectivos complementares. Em primeiro lugar, estuda como as diferentes técnicas de adaptação dinâmica que foram propostas para o modelo de falha por paragem podem ser aplicadas na presença de falhas Bizantinas. Em segundo lugar, através da concretização destas técnicas numa moldura de software comum, baseada no pacote de código aberto BFT-SMaRt, apresenta um estudo comparativo do desempenho das mesmas.

16:20
José Enes (HASLab - INESC TEC & U.Minho, Portugal)
Nuno Machado (HASLab - INESC TEC & U.Minho, Portugal)
Francisco Maia (HASLab - INESC TEC & U.Minho, Portugal)
Miguel Matos (INESC-ID & IST, U.Lisboa, Portugal)
Rui Oliveira (HASLab - INESC TEC & U.Minho, Portugal)
Coerência probabilística em sistemas chave-valor escaláveis
SPEAKER: unknown

ABSTRACT. A escalabilidade é uma das características de maior relevância para os sistemas de armazenamento de dados. Para atingir uma escalabilidade e flexibilidade alta, algumas abordagens relaxam a coerência de dados evitando assim protocolos dispendiosos de coordenação. Estas abordagens são também as que tipicamente lidam melhor com churn e falhas, contudo impedem um modelo mais simples e intuitivo para o cliente, obrigando-o a preocupações iniciais como o nível de concorrência no sistema.

Neste artigo apresentamos um protótipo que junta um sistema de armazenamento de dados capaz de escalar até milhares de nodos com poucas garantias de coerência, com um algoritmo probabilístico de ordem total escalável. Deste modo pretendemos que seja mitigado o problema de coerência de dados, mantendo propriedades de escalabilidade e robustez, oferecendo também ao cliente um modelo simples. Por fim, é proposta uma avaliação deste novo sistema, de maneira a verificar uma compatibilidade entre coerência de dados e desempenho.

16:40
Carlos Morais (Universidade do Minho, Portugal)
José Pereira (Universidade do Minho, Portugal)
Ana Alonso (Universidade do Minho, Portugal)
Impacto de mecanismos de recuperação transacional em sistemas NoSQL
SPEAKER: unknown

ABSTRACT. Com o amadurecimento e larga aceitação das bases de dados NoSQL tem havido um interesse crescente na adição de transações multi-linha, que proporcionem as propriedades ACID sem comprometer o desempenho e capacidade de escala destes sistemas. Apesar das propostas nesta área assentarem em técnicas bem conhecidas de bases de dados, como a multi-versão e a recuperação, a sua aplicação está agora enquadrada em pressupostos diferentes, não sendo claro que os compromissos tradicionais se mantenham.

Neste contexto, este trabalho sistematiza os compromissos relacionados com a escolha de um mecanismo de recuperação, que garante que as alterações efetuadas por uma transação confirmada persistem atomicamente. Além de analisar qual o impacto na arquitetura do sistema da escolha do mecanismo de recuperação, comparamos experimentalmente as alternativas mais interessantes com diferentes cargas de trabalho.

16:45
Pedro Joaquim (INESC-ID, Portugal)
Manuel Bravo (INESC-ID, Portugal)
Rodrigo Rodrigues (INESC-ID, Portugal)
Miguel Matos (INESC-ID, Qatar)
Luis Rodrigues (INESC-ID, Portugal)
Processamento Incremental de Grafos em Sistemas Heterogêneos
SPEAKER: unknown

ABSTRACT. O processamento incremental de grafos tem hoje aplicações nas mais diversas áreas. Tipicamente, este tipo de processamento é feito em infraestruturas de computação na nuvem que oferecem um conjunto bastante diversificado de serviços. A heterogeneidade dos recursos existentes e os diferentes serviços de aluguer de máquinas, reservadas ou efémeras, oferecem custos proporcionais às características pretendidas. Neste artigo apresentamos um sistema de processamento incremental de grafos que pretende usar esta heterogeneidade para reduzir os custos de exploração. A solução proposta realiza a atribuição dos nós do grafo às máquinas usando uma política ciente da heterogeneidade da carga e dos recursos. Esta política é complementada com mecanismos de tolerância a faltas eficientes, para incorporar a utilização de máquinas de baixa fiabilidade e recursos reduzidos sem incorrer em atrasos significativos no processamento.

17:00-17:15Coffee Break
17:15-18:00 Session 11: Encerramento
Chairs:
Helena Rodrigues (Universidade do Minho, Portugal)
António Teixeira (University of Aveiro, Portugal)