Research On Task Offloading And Resource Allocation In Edge Computing Networks

With the rapid development of the Internet of Things applications,the contradiction between the real-time processing requirements of the large-scale computing-intensive tasks and the limited processing capabilities of the terminal devices has become increasingly prominent.By sinking the wireless,computing,and caching resources to the edge of the network and providing computing services close to the source of data,edge computing can effectively alleviate the contradiction as mentioned above.Since the edge computing environment is essentially a wide-area distributed parallel computing environment,and the resource of each edge computing server is limited,it is essential to realize the effective use of scattered and limited resources with the trend of large-scale device connection and explosive data traffic growth in the future.Therefore,the task offloading and resource allocation mechanism is critical to meet the processing delay,processing quality,and execution energy consumption requirements of various tasks.However,due to the complex network environments and resource competition among multiple tasks,unreasonable task offloading decisions can easily lead to imbalanced workload and inefficient resource utilization.Additionally,as task offloading involves the coupling of transmission,computation,and storage processes,the critical performance of edge computing systems can be further restricted by the imbalanced allocation of multidimensional(computing-communication-storage)resources.In order to achieve loadbalanced task scheduling and on-demand optimization of resource allocation strategies,this thesis studies the task offloading and resource allocation optimization mechanism in edge computing networks,which is carried out from three respects,i.e.,edge-side cooperation-based edge computing systems,end-edge cooperation-based edge computing systems,and space-ground cooperation-based edge computing systems.The main contributions of the dissertation include:1)A task offloading and resource allocation optimization scheme is proposed in the edge-side cooperative computing systems.Under the imbalance between the supply and demand of edge server resources,the proposed scheme solves the difficulty of ensuring the mixed performance index requirements of task processing delay and task processing quality.Firstly,leveraging the idea of task scalability,a model for balancing computing quality and resource consumption is introduced to exploit the computational resources fully.On this basis,an optimization problem is formulated to investigate the tradeoff between the service quality and the service delay performance in the edge-side cooperative computing systems.Secondly,to cope with the time-varying computation demands caused by the dynamic wireless transmission conditions,an online algorithm is designed to adaptively optimize the decisions of the access selection,the allocation of the radio resources,and the computation configuration and the data volume selection,which can keep the supply-demand relationship of each edge service's resource in a stable state.In particular,via the variable relaxation and the duality theorem,the complex mixed-integer non-linear optimization problem in each time slot is transformed into DC programming,which can be solved with low complexity.Finally,the experimental results show that the proposed optimization scheme effectively guarantees delay performance and improves the quality of task processing.2)A task offloading and resource allocation optimization scheme is proposed in the endedge cooperation-based edge computing systems.Under the spatial and temporal non-uniformity of the client task arrivals,the proposed scheme solves the problem of high task processing cost caused by an unbalanced network load.Firstly,we consider the service placement problem that the edge nodes place the most up-to-date service program at user devices to enable local task execution at the user side.On this basis,by exploring the differentiation of transmission conditions and server resource competition among time slots,an optimization problem is constructed to minimize the time-averaged service cost.Secondly,to cope with the time-varying network conditions,we use Lyapunov optimization theory to achieve the network workload balance by ensuring network stability.Then,leveraging the combined power of the alternating optimization and the convex optimization methods,we obtain the real-time optimization solution of the data admission control,service model placement,computation,and wireless resource allocation in a coordinated manner,where the load balancing in the spatial domain is realized by guaranteeing the condition of best performance gain.Finally,the simulation results verify that the advantages of geographical location and available resources of each server can be brought into full play,and the proposed algorithm can effectively reduce the cost of task processing.3)A task offloading and resource allocation optimization scheme is proposed to efficiently and collaboratively allocate the multi-dimensional resources in the end-edge collaborative architecture.Under the coexistence of multiple task types,the proposed scheme avoids the high task processing delay caused by the imbalance of multidimensional resource allocation.Firstly,given the impact of task caching deployment on task processing capabilities,the inter-correlation mechanism among bandwidth,computational,and caching resources allocation is deeply considered.An edge caching and computation management problem is formulated to minimize the task processing delay.Secondly,due to the complex decision space and time-varying network state,a distributed online optimization algorithm is designed to solve the formulated problem,in which each user decides the allocation of communication and computing resources according to the local copy of other users.Then,by transforming the request scheduling decision of each user and task cache deployment decision of the edge server into a matching game problem of bilateral utility maximization,each device only needs to rely on local real-time information to make decisions independently.Finally,the performance lower bound of the proposed algorithm and the performance gap compared with the optimal solution are theoretically proved.Furthermore,the simulation results prove that the proposed algorithm reduces the task processing delay within the time-average cost budget constraint.4)A task offloading and resource allocation optimization scheme in the drone-assisted air-ground cooperative system is proposed to meet the challenge of the poor flexibility of the ground edge computing framework and the randomness of the network environments.Firstly,in order to improve the network processing capacity and reduce energy consumption,a three-tier computing offloading framework is constructed to unlock the full potential of the drone,in which the drone roaming around the area may serve as the computing sever to help users compute their tasks or act as a relay for further offloading their computation tasks to the ground server.Secondly,due to the stochastic property of user movements and task arrivals,the task scheduling,the resource allocation,and the trajectory of the drone are jointly optimized in an online manner.Specifically,by utilizing the nonlinear fractional programming and stochastic network optimization,the stochastic optimization problem is decomposed into three separate subproblems in each slot.Then,since there are nonlinear couplings among the variables in the task offloading and trajectory design problem,we first drive the optimization solution of the task offloading decisions in closed forms and design an iterative algorithm to achieve the approximate optimal solution of the trajectory design.Finally,simulation results verify that the proposed algorithm can effectively improve the long-term time average energy efficiency performance without relying on the prior knowledge about the future network state information.