Research On Computation Offloading And Resource Allocation In Service Caching-assisted Edge Network

In 5G communication system,the access of a large number of smart devices and the proliferation of emerging services lead to an explosion of data traffic load and terminal resource pressure.Centralized cloud computing brings high backhaul overhead and transmission delay,which cannot meet the increasingly stringent requirements of users for network performance and quality of service.Meanwhile,repeated requests for a large amount of popular data generate data redundancy,causing the waste of network resources and energy.Service cachingassisted edge computation offloading caches the service program data required for task execution at the edge of network,which allows tasks for offloading computing and provides tasks with nearby and reusable service program data by fully utilizing edge storage,computation and communication resources.It has been considered as an important computing paradigm to realize ultra-low latency and low power green communication in the future wireless network.In this paper,joint optimization of service caching placement,task computation offloading decision and multi-dimensional resource allocation is studied to reduce the cost of task execution in edge networks.The main work of this thesis are as follows:(1)In the edge computing scenario with multi-task users,based on splitting required data for task computing into input data and service program data,a service-oriented service caching-computing model is proposed to allow multiple tasks to reuse cached service program data and characterize the heterogeneity of tasks.In order to further utilize distributed resources on terminals,users are clustered into cooperative group and device-to-device technology is combined to realize nearby transmission and sharing of cached data among users.Aiming at minimizing caching benefit ratio-based system cost of weighted users’energy consumption and execution delay,service caching,task offloading and spectrumcomputing resource allocation schmems are jointly optimized.The original optimization problem is decomposed into three subproblems and solved iteratively.Simulation results imply that compared with other baselines,the proposed joint optimization algorithm can achieve better caching gain and lower system cost,and the weighted cost model can realize tradeoff between execution delay and users’ energy consumption through dynamic weight control.(2)In order to further enhance the cooperative interconnection and delay service quality guarantee in the edge network,deterministic computing power network is combined to provide deterministic computing power for tasks by resource reservation scheme,and hybrid distributed-centralized resource pools are established to support cloud-edge-end cooperative vertical computing and horizontal caching among multi-edge servers based on the aforementioned proposed service caching-computing model.Moreover,non-orthogonal multiple access is introduced in the edge multi-small cell to enhance the network capacity.Based on local popularity-caching diversity,hypergraph coloring and geometric programming theories,the joint optimization algorithm of service caching placement,task computing power matching and communication resource allocation is developed to minimize the total system delay.Simulation results reveal that the proposed scheme can acquire lower total sysytem delay and remarkable cooperative caching gain with meeting the guarantee of delay,and it has better network performance compared with orthogonal multiple access system.