Dynamic adaptation of interaction models for stateful web services

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

Wireless Sensor Networks (WSNs) are accepted as one of the fundamental technologies for current and future science in all domains, where WSNs formed from either static or mobile sensor devices allow a low cost high-resolution sensing of the environment. Such opens the possibility of developing new kinds of crucial applications or providing more accurate data to more traditional ones. For instance, examples may range from large-scale WSNs deployed on oceans contributing to weather prediction simulations; to high number of diverse Sensor devices deployed over a geographical area at different heights from the ground for collecting more accurate data for cyclic wildfire spread simulations; or to networks of mobile phone devices contributing to urban traffic management via Participatory Sensing applications. In order to simplify data access, network parameterisation, and WSNs aggregation, WSNs have been integrated in Web environments, namely through high level standard interfaces like Web services. However, the typical interface access usually supports a restricted number of interaction models and the available mechanisms for their run-time adaptation are still scarce. Nevertheless, applications demand a richer and more flexible control on interface accesses – e.g. such accesses may depend on contextual information and, consequently, may evolve in time. Additionally, Web services have become increasingly popular in the latest years, and their usage led to the need of aggregating and coordinating them and also to represent state in between Web services invocations. Current standard composition languages for Web services (wsbpel,wsci,bpml) deal with the traditional forms of service aggregation and coordination, while WS-Resource framework (wsrf) deals with accessing services pertaining state concerns (relating both executing applications and the runtime environment). Subjacent to the notion of service coordination is the need to capture dependencies among them (through the workflow concept, for instance), reuse common interaction models, e.g. embodied in common behavioural Patterns like Client/Server, Publish/- Subscriber, Stream, and respond to dynamic events in the system (novel user requests, service failures, etc.). Dynamic adaptation, in particular, is a pressing requirement for current service-based systems due to the increasing trend on XaaS ("everything as a service") which promises to reduce costs on application development and infrastructure support, as is already apparent in the Cloud computing domain. Therefore, the self-adaptive (or dynamic/adaptive) systems present themselves as a solution to the above concerns. However, since they comprise a vast area, this thesis only focus on self-adaptive software. Concretely, we propose a novel model for dynamic interactions, in particular with Stateful Web Services, i.e. services interfacing continued activities. The solution consists on a middleware prototype based on pattern abstractions which may be able to provide (novel) richer interaction models and a few structured dynamic adaptation mechanisms, which are captured in the context of a "Session" abstraction. The middleware was implemented and uses a pre-existent framework supporting Web enabled access to WSNs, and some evaluation scenarios were tested in this setting. Namely, this area was chosen as the application domain that contextualizes this work as it contributes to the development of increasingly important applications needing highresolution and low cost sensing of environment. The result is a novel way to specify richer and dynamic modes of accessing and acquiring data generated by WSNs.

Este trabalho foi parcialmente financiado pelo Centro de Informática e Tecnologias da Informação (CITI), e pela Fundação para a Ciência e a Tecnologia (FCT / MCTES) em projectos de investigação