Resource Scheduling Strategies And Optimization Techniques Supporting Cloud-Fog-Thing Integration

In recent years,the maturing cloud computing technology has caused a rapid expansion in the number of data centers,the accompanying problems of low utilization rate of platform resource has gradually acquired a lot of extensive attention from both academia and industry.It has become a key issue for the sustainable development of cloud computing industry that how to effectively improve resource utilization and platform throughput.With the development of Internet of things and the growing number of smart mobile devices,the characteristics of smart mobile devices,such as environmental instability,real-time tasks and diversified requirements,pose a more severe challenge to traditional cloud computing technology.As a new emerging service paradigm,fog computing,with the characteristics of context-aware distributed computing,low task response delay and high service quality,has effectively made up for the shortcomings of traditional cloud computing technology.Though cloud computing technology and fog computing technology have got abundant research achievements,the diverse factors of decentralization of data resources,uncertainty of service quality and diversification application demand of users bring new challenges to the realization of efficient and reliable computing services.For example,in current research,there is a lack of consistent resource scheduling framework for cloud-fog-thing hierarchical architecture providing technical reference for services computing.Besides,a more efficient network abstraction and corresponding resource scheduling method are required to tackle performance bottlenecks problems of network data streaming transmission in cloud environment.Aiming at the challenges faced by computing services in mobility-aware and resource scheduling consistency optmization in current cloud-fog-thing hierarchical architecture,we conduct a series of research work on resource scheduling strategies and optimization techniques over cloud-fog-thing hierarchical architecture.Specifically speaking,the contributions of this dissertation mainly includes the followings.(1)In order to effectively solve the problem of resource fragmentation and distribution imbalance in the process of serving users in the cloud and fog environment,a consistency optimization framework for efficient integration and optimization of resource scheduling in cloud-fog-thing hierarchical architecture is proposed in this dissertation.This resource scheduling framework integrates resource by leveraging software-defined networking and network functions virtualization technology.Basing on distributed computing service architecture,this framework integrates the computing service response module,the computing service allocation module and the computing service execution optimization module.Specifically speaking,the framework divided into three layers from bottom to top,that is terminal access layer,fog computing layer and cloud computing layer.The terminal access layer collects computing service requests from users through unified computing service interface to achieve service consistency access for wired and wireless users.The fog computing layer extends cloud-based services to the network edge and selects the service channels according to the requirements of the actual computing services of the users.The fog computing layer is responsible for the time-sensitive applications,and it could support mobility-aware resource scheduling.The cloud computing layer is responsible for the big data processing which transferred from fog computing layer,and it schedules tasks that consume huge storage resources and computing resources based on the traditional large-scale parallel processing data framework over the cloud platform.(2)In order to tackle the problem of low network utilization during the big data processing and optimize the network resource utilization in cloud environment,this dissertation proposes a deadline-aware network resource scheduling optimization method for cloud computing platform.To be specific,this dissertation begins with the network data stream transmission processing in the big data analysis processing in cloud environment,which is often the bottleneck of the whole analysis processing job's completion.Then,the generation process of the network data streams is analyzed,and the collaborative network flow(Coflow)abstraction is introduced to describe the semantic relations between network data streams in big data processing.Further,the placement constraints of network data streams are analyzed based on Coflow abstractions.Considering the impact of bandwidth allocation strategy of computational node on the completion time of big data processing,the priority of network data stream transmission of each computing node in big data processing is determined in the proposed scheduling method.The proposed scheduling method could effectively shorten the average completion time of time-sensitive jobs,and improve the throughput of cloud platform.(3)In order to effectively solve the problem of resource scheduling imbalance faced by the fog computing platform when processing multi-user computing service requests.This dissertation proposes a platform-oriented optimization method which satisfies the task dependency constraints for resource scheduling to improve the utilization of system resources.To be more specific,through analyzing the spatio-temporal information of multi-user computing service requests and the heterogeneity of the configurations of fog computing service nodes,we introduce the placement constraints of computing tasks on service nodes and take the dependency constraints of tasks into consideration,and then the scheduling problem of computing tasks is modeled and analyzed.We put forward to use job's slack time to quantify the system resource utilization index in the multi-user fog computing platform,and based on this index,we further formalize the problem of improving the system resource utilization as an optimization problem that minimizes the sum of the slack time of all jobs,and presents the problem's NP-Hard property.Considering the fact that the problem is difficult to handle,an offline scheduling algorithm based on binary relaxation technology with theoretical basis and online scheduling algorithm based on list scheduling method are given respectively.The proposed algorithm could reduce the computing service response time and improve the system resource utilization.(4)In fog computing environment,considering the fact that there are high demands for security and efficient management of service nodes in addition to the mobility support for providing computation offloading service,this dissertation brings blockchain technology to manage the geographically dispersed fog computing service nodes,and therefore provide users with the technical support of fully distributed real-time secure computing offloading service.Furthermore,the BMO(Blockchain-based Mobility-aware Offloading)method is proposed for the scheduling of tasks.Specifically speaking,BMO divides the mobile characteristics of users into those with path constraints and without path constraints.For the mobile scheme with path constraints,BMO firstly constructs the R-tree index for the service coverage of each computing service node,and then selects the offloading service node according to the service expenditure and trajectory prediction of mobile users based on the deadline of the task.As for mobile scheme without path constraints,the BMO method predicts the dwell time of users in the range of offloading service by introducing individual mobility model,and optimizes the selection of the available service nodes according to deadline and service expenditure of the task.BMO method not only do the best to meet deadline of computational task through offloading technology,but also reduce the offloading service expenditure of task through optimized selection of offloading service nodes.