Research On Online Dispatching And Fair Scheduling Of Edge Computing Tasks

Due to the higher requirement of information processing speed in modern warfare,the military has built many large cloud data centers to cope with the growing demands of task processing.However,these large data centers are often far away from end devices,resulting in long transmission delays.The emergence of edge computing can effectively solve the problem of long-distance transmission delay between user devices and remote cloud servers.Users can dispatch tasks to nearby edge servers for calculation and minimize the average response time of tasks through efficient task scheduling methods.However,in the task dispatching phase,the dynamic nature of the network state and server load makes it difficult to select the optimal edge server to offload the task.In the task scheduling phase,each edge server may face a large number of offloaded tasks to deal with,resulting in longer average task response time and even serious task hunger situations.In addition,a large number of edge servers also make the high total power consumption.Combining with the latest online learning and reinforcement learning method,we proposed edge computing tasks online dispatching and fair scheduling algorithm,which mainly includes the following three aspects: 1)We propose a delay-aware task dispatching method based on online learning,which can estimate the network status in real-time and the load on the server,so as to dispatch the task to the optimal edge server.2)We propose a delay-aware task scheduling method based on reinforcement learning.By combining with the traditional round-robin algorithm,it can achieve both efficiency and fairness in the scheduling process.3)We propose an energy-aware task dispatching method based on online learning,which can obtain real-time information by interacting with the environment,so as to minimize the overall energy consumption while ensuring acceptable Qo S.Experimental results show that the proposed method can dynamically allocate network resources and computing resources to the offloaded task according to their time sensitivity and energy consumption requirements.Compared to the existing methods,the proposed method could effectively reduce the average response time and ensure fairness,which proves its effectiveness and high efficiency.