Research On Offloading Strategies For Dependence Tasks Based On Cloud And Edge Collaboration

Limited by terminal performance,it is difficult for computing-intensive and delay-sensitive applications to run stably on mobile terminals.The offloading by sending data to cloud servers with powerful computing capabilities,has become an effective means to reduce the burden on mobile devices.However,considering network congestion and limited backhaul links,it is difficult to meet the low-latency service requirements of computationally intensive applications using mobile cloud computing(Mobile Cloud Computation,MCC).Therefore,low-latency and resource-limited mobile edge computing(Mobile Edge Computation,MEC)complements the MCC.This thesis studies the intelligent offloading problem of cloud and edge collaborative computing based on dependence tasks,and the main contributions are as follows.Firstly,a Collaborative Cloud-Edge Computing Offloading(CCECO)algorithm is proposed for single-thread dependence tasks.For a single-thread dependent task,the problem of minimizing the energy consumption under a delay constraint is transformed to a mixed-integer nonlinear programming.The problem is divided into two parts: integer part and non-integer part.The integer programming problem is an NP-hard problem.We propose a one-time offloading principle to convert the NP-hard problem to a linear time complexity problem based on the hypothesis that cloud-edge-end computing power decreases.The non-integer part is a convex optimization problem.We obtain the optimal solution according to the KKT condition and the stochastic gradient descent method.Finally,the thesis presents a simulation analysis of the above-studied problem Simulation results show that the proposed algorithm can effectively reduce the energy consumption at the cost of low complexity.Secondly,as multi-threading greatly increases the decision space dimension,this thesis proposes the critical path to reduce the decision space dimension.We mark the thread with the largest computational task in the parallel computing part as the critical path,and the computational latency of the remaining sub-threads in the parallel computing part should be consistent with the critical path.On the basis of this scheme,a CCECO algorithm for multi-thread dependence tasks is proposed,which uses the one-time offloading principle and gradient descent to obtain the optimal solution for the integer part and the non-integer part.Finally,the CCECO algorithm is analyzed,and the numerical results show that the strategy is effective in reducing the computational offload energy consumption of multi-thread dependence tasks.Moreover,the adaptability of the proposed strategy can also automatically adjust the offloading strategy of each thread,so as to ensure the minimum energy consumption.