Software Adaptability Study For Mobile Environments In Pervasive Computing

Mobility and easy use are the two main concerns of current society on IT technology. People are expecting to be able to quickly benefit from a variety of background services existed all the time no matter where he/she goes. The two requirements not only propel rapid development of wireless technology and wide acceptance of mobile devices, but also deviate the traditional computing style from a central controlled, well organized way to an ad hoc dominated pervasive computing. Compared to hardware and network support, the software development for pervasive computing is far behind, which in return, limited the application and deployment of pervasive computing to some extents.Unlike the traditional distributed computing software running in stable environments with rich resources, pervasive computing software runs in unforeseen wireless network environments with mobile users, limited resources and continuously altered context. These put a great deal of demands on the software adaptability.From perspectives of mobility, including user mobility, device mobility and code mobility, this thesis mainly concentrates on the software adaptability issues from configuration to runtime in both infrastructure and ad hoc networks environments. In order to better serve individual users and minimize the resource requirements in a dynamic environment, multiple technologies, such as source routing, mobile code, rule mechanism, policy language and graph theory are further explored and used to study problems in service discovery, service composition, optimized selection of design paradigms and optimized service allocation. The major contributions of this dissertation are:(1) Based on DSR routing protocol, an active service discovery and composition protocol for mobile ad hoc networks (MANETs), called SC-QoS, is presented to meet Quality of Service (QoS) and conserve resources of mobile nodes. The dynamic source routing and cross layer design are used to transform the problem of a service discovery and composition in a MANET into that of a service path discovery in a service network. With a single process on service path discovery, it can handle route discovery, service discovery and service composition simultaneously to meet QoS requirements and decrease control overhead. Simulation results show that the proposed protocol can employ a relatively small number of control messages to obtain high success rate and conserve more resources by using the constrained flooding method and local path maintenance mechanism.(2) The qualitative and quantitative approaches to design paradigm selection based on mobile code are proposed. After analyzing the features of six types of design paradigms, including MONO, C/S, COD, REV, MA and MC, a qualitative selection method of design paradigms considering software mobility and context is introduced to satisfy user requests in different contexts. The rules and policy language are used to separate paradigm selection codes from function codes, which can realize dynamic selection of paradigms and ease system maintenance. In order to conserve limited resources, a quantitative selection algorithm for design paradigms is presented by optimizing resource consumption.(3) A novel algorithm for the optimal software allocation is presented to comply with component mobility constraints and conserve resources of mobile devices. Multiple mobility constraint relationships, such as component mobility, local invocation, remote invocation, replication and rebinding, are thoroughly used to conserve limited resources. Based on network flow theory, the problem of an optimal software allocation is reduced to a minimum cut problem of the flow network using equivalent transformation rules. Then the max-flow min-cut theorem is used to calculate the optimal allocation of an application between two hosts and the suboptimal results among multiple hosts. To achieve the algorithm's generality, an abstract cost graph model is set up to save multiple resources, such as network bandwidth usage, response time and energy consumption, respectively and simultaneously. Experimental results demonstrate that the proposed partitioning algorithm for the constrained cost graph not only comply with software properties and save more resources, but also can well adapt to environmental changes.(4) An adaptive prototype system in pervasive computing environments has been implemented. With the aid of the Jist/Swans network simulator, the SC-QoS protocol for MANETs is implemented based on the DSR protocol. Multiple design paradigms are implemented on the JADE mobile agent platform. Finally, a generic algorithm for quantitative computation and optimized allocation is implemented to suit different design paradigms and different resources.This thesis focuses on software adaptability in mobile environments. During software configuration and runtime, four correlative adaptive technologies with the same goals are integrated within the middleware. Finally, simulation experiments and the prototype system have demonstrated the validity and practicability of the proposed theories and algorithms.