Mapping And Implementation Of Dispersed Computing Paradigm Based On Path Computing

The emergence of 5G communication technology has brought unprecedented development opportunities for the diversified IoT networks,for which many Internet of Everything(IoE)elements have changed from conceptual imagination to reality.At the same time,a large number of smart devices are connected to the Internet and the growth of the IoT applications analysis data is also increasing.At present,cloud computing is the main tool for the IoT.To some extent,it solves the problems of high performance business computing and mass analysis data storage in the IoT applications.However,fast and reliable business processing becomes critical in the future filled with the IoT devices data driven by 5G.Cloud computing has been unable to satisfy the demands of the emerging IoT applications for real-time,agility,reliability and security.Because it still has the disadvantages of high latency,limited communication resources,and single point of failure faced with special services.The relevant theory and implementation technology of the dispersed computing paradigm is researched in the thesis in order to solve the above-mentioned disadvantages when cloud computing is applied to the future IoT networks.It uses the computing and storage resources that are close to the users and dispersal across the domain to perform business computing in a collaborative way and improve the applications performance of the business carrier networks based on the dispersed computing architecture which provides users with realtime and agile computing services in a collaborative and shared way.In this thesis,the concept of path computing,one of the realization methods of dispersed computing,is expounded and its task mapping strategy is researched.At the same time,a dispersed computing hardware simulation platform based on path computing is designed and implemented.The dispersed computing paradigm is explored in the thesis from the perspectives of theory and engineering.The main research contents are as follows:The overview on present research situation and related theoretical and technical foundations of the thesis are explained in order from the three aspects of distributed computing paradigm,path computing strategy and container virtualization technology,which lays the theoretical and technical foundations for the following researches on the tasks mapping strategy of path computing and the implementation of dispersed computing hardware simulation platform.A cloud and fog network architecture is designed based on the existing hospital infrastructures to support the realization of path computing paradigm in order to solve the problem of high latency in business response when cloud computing is applied to the medical big data processing.And a path computing task mapping strategy is studied based on the network architecture,which defines a mapping rule beetween the big data task model in the form of a directed acyclic graph and the hospital network topology graph in the form of an undirected graph.According to the above mapping rule,a latency optimization scheme based on the discrete binary particle swarm optimization algorithm is proposed,and an experimental simulation is performed with the Matlab platform.Simulation results and a comparison with the cloud computing show that the proposed path computing scheme reduces the latency by more than 50% when the amount of data is 5-10 Mb.A dispersed computing hardware simulation platform architecture is designed based on the research of the path computing task mapping strategy.The network simulation module that can automatically generate a satellite cluster simulation network,the subtask computing modules for each category in the task graph and the resources pool of the computing modules in the architecture are sequentially implemented based on the Docker container and its network and orchestration technology.It uses the platform to perform simulation task graph of image processing,verifying the feasibility of dispersed computing paradigm based on path computing and the latency performance advantage of task graph mapping strategy based on discrete binary particle swarm optimization algorithm by the physical equipments.Simulation results and a comparison with the cloud computing show that the path computing scheme reduces the latency by more than 40% when the amount of image data is 4 Mb.