多任务边缘计算中最小化时间的资源分配方法

Resource allocation method for minimizing time in multi-task edge computing

针对边缘计算系统中多个计算任务之间存在某种依赖关系的特点,研究一种最小化总计算时间的资源分配策略。考虑多个任务之间的顺序依赖关系,用户的多个任务按顺序依次卸载;在当前任务卸载完成时,不用等该任务完成计算,就开始卸载下一个任务。通过引入一种两层卸载策略,用户可以先将任务卸载到小基站边缘服务器,当小基站边缘服务器计算能力不足时,小基站再将部分任务卸载到宏基站边缘服务器。建立联合优化用户关联、计算资源和用户发射功率的资源分配问题,达到最小化总计算时间的目标。采用量子行为粒子群优化算法进行求解,得到全局次优解。仿真结果表明,与标准粒子群优化算法和其他基准策略相比,使用量子行为粒子群优化算法所得到的总计算时间更少。

Aiming at the characteristic of a certain dependency relationship existing among multiple tasks in the edge computing system,a resource allocation strategy for minimizing the total computing time is investigated in this paper.Sequential dependency relationship among multiple tasks is taken into account.Multiple tasks of the user are offloaded in sequence.When the current task completes offloading,the next task can be offloaded without waiting for the current task to finish computing.By using a two-tier offloading strategy,user can first offload task to small base station(SBS),and when the edge server in SBS has insufficient computing capacity,SBS will offload the part of task to the edge server in macro base station.The joint optimization of user association,resource allocation of computation resources and the transmitting power of user are formulated to minimize the total computation time of the multi-task edge computing(MEC) system.A suboptimal solution is obtained by adopting a quantum-behaved particle swarm optimization(QPSO) algorithm.Simulation results show that the QPSO algorithm has less total computation time compared with the standard particle swarm optimization algorithm and the other benchmark strategies.