Dynamic Software Updating In Pervasve Computing

Nowadays a wide set of portable digital devices (e.g. cell phones, laptops, digital cameras, Wi-Fi devices, etc) and miniaturized devices (e.g. sensors) penetrate in modern cities deeper and deeper. Pervasive computing plays an important role in our daily life, and wishes the services are available anywhere and every time. In order to turn the vision into reality, the systems deployed in pervasive environments need ability of dynamic updating to enhance the availability of services. Dynamic updating is a kind of software evolution technology that updates software modules without interrupting the running system. So it can help to decrease the downtime and avoid the performance downgrade of a system during the period of maintenance. However, the characteristics of pervasive computing, sush as decentralization and heterogeneity, bring new challenges to the design and implementation of dynamic updating.This thesis discusses how to implement dynamic software updating in pervasive computing, and is dedicated to provide dynamic updating extension to an existing service platform in a non-invasive way. The main contents of this thesis contains the following parts:(1) According to the characteristics of pervasive computing, this thesis sets several goals for the approach. It investigates the mechanism of dynamic updating, points out the shortage of centralized architecture and proposes a decentralized architecture .(2) This thesis uses the process algebra FSP to formalize the mechanism and check the formal model against some certain properties.(3) Based on the service platform OSGi/R-OSGi, the updating model is put into practice and an architecture for updating framework is designed. This thesis also introduces a four-stage updating procedure and discusses the key issues in an implementation of dynamic updating, which contains the granularity of updated units, update timing, state transfer, dependency resolution, error tolerance.(4) A prototype system is developed. Based on the system design and experiments the whole approach is evaluated, and the conclusion shows this approach match the requirements.