| In recent years,with the rapid development of the Internet of Things,cloud computing,artificial intelligence and other technological fields,digital transformation and the Internet of Everything are gradually pushing the information age into a new era.The massive data generated by 5G large-scale access scenarios requires continuous optimization of the network and improvement of data processing capabilities to respond to various types of applications,which puts forward higher technical requirements for updating the current computing model.Serverless computing provides function-level functional services in an event-triggered manner,and its pay-per-use method can achieve efficient resource utilization.This research was initially born for cloud computing and gradually expanded to the network’s edge.Combining the efficient resource management of serverless computing and the short-distance transmission of edge computing,serverless edge computing provides a theoretically perfect solution for periodically triggered Io T applications.Model.However,while the paradigm shift brings new opportunities for Io T applications,it also brings new problems.This thesis focuses on UAV services with great potential in the future.It constructs a distributed UAV self-organizing communication network equipped with FaaS services,which can achieve flexible deployment and scheduling to provide high-quality computing services.The main research contents are summarized as follows:1.In view of the cold start delay and resource utilization efficiency of FaaS services,consider the container cach e strategy of joint request distribution and cache FaaS services in the form of containers to UAV nodes to optimize the cold start delay on the responsiveness of Io T services impact,while taking into account the efficiency of cache resources.Firstly,it is revealed that the caching decision can be mapped to the classic ski rental problem.The problem is gradually mapped to the multi-store ski rental problem from the single-UAV serial request arrival mode to the multi-UAV concurrent request mode.If the prior information of request arrival is unknown,the cache decision is realized based on a set of probability distribution functions that combine node selection and cache duration.On this basis,the online algorithm(FaaSCC)is further considered under reso urce constraints and communication delay.This algorithm distributes requests and makes cache decisions by balancing the load of each node.Finally,the request arrival simulation experiment verifies the effectiveness of the FaaSCC algorithm on the overall system cost,which is significantly improved compared with the primary method.2.Aiming at the durability of the UAV network,it is considered to realize the efficient utilization of UAV system energy through UAV deployment and scheduling optimization.At the same time,to further improve UAVs’ endurance,the replacement energy of fixed charging facilities is used.First,this thesis designs a 3D deployment scheme for UAVs based on the K-means algorithm,which can dynamically deploy UAVs and adjust the number adaptively according to user location information.Among them,under the constraint of the maximum tolerable transmission delay,a variable-height UAV capacity model is established,and UAV utilization is defined to evaluate the deployment performance.Secondly,this thesis constructs a 3D energy consumption model of UAVs and considers energy-saving scheduling of UAV groups in a centralized management and control manner through energy monitoring.Finally,the scheme’s feasibility is verified through simulation experiments,and it has a noticeable improvement in optimizing system energy consumption. |
