| The Internet of Things consisting of a large number of terminal devices provides the basis for complex task execution and multi-user concurrent requests under edge computing.However,because the Internet of Things is mostly composed of battery-powered terminal equipment,and some of the network structure is complex and the deployment environment is harsh,it is difficult to replace the battery when the terminal equipment is exhausted.Therefore,the problem of terminal equipment energy consumption optimization has long been studied by academia and industry.The limited computing and storage capabilities of Io T terminal devices are difficult to meet the long-term,complex,and intensive requests of multiple users;edge servers have strong computing and storage capabilities,but are also limited by resources;cloud servers have powerful resources and performance,but it is far away from users,and it is difficult to meet the data transmission requirements of delay-sensitive application scenarios.Therefore,the deployment of a reasonable cloud-edge-end collaboration strategy provides the effective method to optimize the energy consumption of the Internet of Things.The caching mechanism is an effective energy optimization method.In certain scenarios,the terminal data of the Internet of Things exhibits a certain degree of stability.Encoding historical data with erasure coding and co-caching,combined with data prediction and inspection to ensure the validity of the data,can ultimately reduce the energy consumption of network data transmission and improve the quality of user service.The main research contents of this thesis are as follows:Firstly,this thesis uses multi-user concurrent request optimization technology,which is to code service requests with the same or similar temporal and spatial attributes to generate request clusters.A request cluster contains two or more user public service requests,and has the same time and space attribute information.Using this method can reduce the transmission of redundant request information and service data,and optimize network energy consumption.Secondly,build a cloud-edge-end collaborative caching mechanism based on erasure coding.After obtaining the terminal data and completing the service response,the data is first cached in the edge server.When the storage capacity of the edge server cannot meet the storage requirement,the cached data is encoded by erasure coding to generate a cache block,and the cache block contains a specific number of data blocks and check blocks.Then the edge server gradually distributes the cache blocks to the Io T terminal devices and caches them in a manner of increasing the number of communication hops.When the data needs to be cached to respond to a user's specific request at a certain moment,the edge server and the terminal device cooperate to complete the cache assembly and decoding.Thirdly,train a data prediction model based on the Long Short-Term Memory(LSTM)neural network,and build a cache validity check and update mechanism.Deploy an accurate historical data-based prediction model on a cloud with powerful performance,and deliver the trained model to the edge server.The validity of the cached data is weighted by the difference between the cached data and the predicted data on the edge server and the cached duration of the cached data.When the new real data is acquired,it will also be compared with the predicted data.If the prediction deviation exceeds the threshold,the correction signal and real data will be sent to the cloud server to correct the forecast model.Finally,the simulation experiment method is used to verify the effectiveness of the cloudedge-end collaborative caching strategy constructed in this thesis.The results show that the algorithm in this thesis reduces the total energy consumption by 49.3% compared with the public request offloading algorithm,and optimizes the total energy consumption by 20.9% compared with the similar collaborative caching algorithm. |
PDF Full Text Request