Mobile collaborative cloudless computing

Tese de doutoramento, Informática (Engenharia Informática), Universidade de Lisboa, Faculdade de Ciências, 2015Although the computational power of mobile devices has been increasing, it is still not enough for some classes of applications. In the present, these applications delegate the computing power burden on servers located on the Internet. This model assumes an always-on Internet connectivity and implies a non-negligible latency. Cloud computing is an innovative computing paradigm where the resources made available by a number of servers are transparently shared by its users. Cloud computing simplifies resource management, establishing the ground for the elastic computing concept, where each user can easily enlarge or reduce his amount of resources. The thesis studies the challenges and contributions posed to the application of the cloud computing concept to wireless networks. The goal is to define a reference architecture for high performance mobile application, the Collaborative Cloudless Computing (C3) together with a framework that addresses the challenges raised decomposing it on a series of components. The framework, named Mobile Collaborative Cloudless Computing Environment (MC3E) dismisses the connection to the Internet. In this scenario, computing power is obtained from neighbouring mobile devices, which coordinate to achieve a common goal: the execution of tasks requested by one or more participants. Expectations are that the use of the C3 concept contributes to: i) reduce the response time, in comparison with delegations on commercial clouds; ii) reduces user frustration when Internet connectivity is not available or its bandwidth is not sufficient; and iii) alleviates the bandwidth consumed by these applications in the cellular infra-structure. Allowing a mobile device to provide a service to the neighbouring peers carries non-negligible risks, of which confidentiality, privacy and selfishness are good examples. To discourage selfishness, two approaches are typically taken: i) in trade based systems, devices agree on a retribution for the execution of a task; ii) in trust based systems rogue devices are detected and advertised by their peers. This thesis describes and evaluates a hybrid system, combining trade and trust-based characteristics. We call this approach the Hybrid Trust and Trade system (HTnT). HTnT suits well privacy requirements as it assumes and encourages users to frequently change their pseudonyms. The work shows that the service detects several misbehaving approaches, even without requiring interacting devices to be connected to the Internet. HTnT will use the capability to anticipate a contact with another device. This knowledge is useful for other applications that rely on some form of data harvesting or hoarding. One of the most promising approaches for contact prediction is to extrapolate from past experiences. This work investigates the recurring contact patterns observed between groups of devices using an 9-year dataset of wireless access logs, produced by 76479 devices that connected to one of the 239 access points of the eduroam network at the Lisbon Polytechnic Institute (IPL). This effort permitted to model the probabilities of occurrence of a contact at a predefined date between groups of devices using a power law distribution that varies according to neighbourhood size and recurrence period. In the general case, the model can be used by applications that need to disseminate large datasets by groups of devices. As an example, we present and evaluate an algorithm that provides daily contact predictions, based on the history of past pairwise contacts and their duration, that will be applied into HTnT. Human mobility pattern analysis also used the same dataset. The understanding of human mobility patterns is key for the development and evaluation of ubiquitous applications. To circumvent the scarcity and difficulties in capturing mobility data, a number of models has been devised. The accuracy in replicating observed human mobility by these models varies. In general, each model concentrates n replicating some of the metrics that have been observed, while neglecting others. Unfortunately, all tend to neglect diversity, in the roles and goals of the users but also in the devices that are used to access the wireless network. We present MobIPLity, a mobility scenario generator that extracts mobility traces from the access records of the IPL dataset. MobIPLity is made publicly available in the expectation that its large scale permits to support evaluations based exclusively on real mobility data, thus removing the uncertainty that emerges from the use of synthetic mobility models. Traces emphasise the differences that can be found between device types, with impact on aspects like the observed trace duration, speed, pause times, inter contact times and availability and which can hardly be replicated on synthetic mobility models. The extracted mobility traces allowed for a comparison with other mobility models, where it was observable that the increasing number of smartphones resulted in significant changes to the utilization pattern, with impact on the amount of traffic and users connection time.Large-Scale Informatics Systems Laboratory (LaSIGE