Resource Management In Fog-assisted Cloud Computing For Internet Of Things

With the rapid promotion and application of Internet of Things(IoT)ser-vices such as smart city and intelligent manufacturing,the decentralization of IoT devices and the centralization of decision-making processes lead to a large amount of data moving in the execution environment,increasing the consump-tion of bandwidth resources.Fog-assisted cloud computing can provide low latency,low energy consumption,and highly reliable computing power support for IoT services.However,during the execution of IoT services,the massive-ness,fragmentation,and randomness of data as well as the dynamic network environment have brought unprecedented challenges to the optimal manage-ment of various resources that provide supporting services.In a distributed fog-assisted cloud computing environment,it is of great research significance to address resource management and scheduling related to IoT services from the perspective of reducing resource consumption.This thesis studies how to effectively reduce the transmission and comput-ing overhead of massive IoT service data when ensuring quality of service.It makes research from three respects,i.e.,distributed execution of large-scale IoT service processes,energy-efficient resource management of cloud computing and fog-assisted cloud computing when processing dynamic IoT service data.The main contributions of this thesis are summarized as follows.Firstly,this thesis studies the fragmentation and distributed deployment algorithms of IoT service processes in fog-assisted cloud computing environ-ment to reduce the amount of data movement in the communication network.Specifically,an IoT service process is abstracted as a directed graph.Based on the breadth first traversal strategy and the spatial attribute analysis of process elements,this thesis proposes a fragmentation algorithm of IoT service process based on spatial attributes.In addition,considering the fact that the empirical data with a certain level of uncertainty are used to guide the optimal deployment of sub-service processes,the uncertain programming theory is used to trans-form the deployment problem with uncertainty into an equivalent deterministic allocation problem,and this thesis proposes a distributed optimal deployment algorithm of sub-service processes that can meet a given confidence level.Secondly,aiming at the energy-efficient resource management problem faced by the cloud in fog-assisted cloud computing environment,this thesis proposes a distributed online optimization approach.Through time decou-pling,the complex resource management problem in multi-core parallel cloud computing is decoupled into CPU time allocation and voltage configuration of processor cores,the data dispatch and scheduling between servers and between different processor cores within a server,which are optimized simultaneously to maximize the energy efficiency of the cloud computing center.It is also proved that the non-convex core configuration problem can increasingly exhibit convexity and achieve the global optimum with growing workloads,maintain-ing the asymptotic optimality of the entire cloud computing center.In the case that the cloud is lightly loaded,the proposed approach dramatically reduces the energy consumption with a marginal loss of throughput.In the case that the cloud is heavily loaded,the approach efficiently translates the power usage tothe processing capacity and hence throughput.Finally,aiming at the energy-efficient resource management problem in fog-coordinated cloud computing when processing massive and fragmented IoT service data,this thesis adopts the Map Reduce idea to conduct joint mod-eling of computing nodes in distributed systems,and proposes a distributed online optimization approach.Through time decoupling,the complex resource management problem in the fusion of edge computing and cloud computing is decoupled into raw data dispatch,result data aggregation and computing resource configuration,which are simultaneously optimized to achieve the life cycle management of IoT service data in an efficient and low resource con-sumption manner.Furthermore,the fog-coordinated cloud computing model is extended to the fog-assisted cloud computing model,i.e.,besides routing and forwarding data,the fog nodes also process part of data.Based on the theoretical analysis,the critical condition,under which the fog-assisted cloud computing can outperform its fog-coordinated counterpart in terms of energy efficiency and data processing delay,is analytically identified.