Research On Task Offloading Online Algorithm For Data Stream Edge Computing

Recently,the number of Internet of Things(IoT)applications has been growing exponentially,and they are generally time-sensitive and resource-hungry applications,such as self-driving and face recognition.Due to the physical size of IoT devices,the computational resources need to be dynamically managed to cope with the real-time tasks.To solve this kind of problems,Edge Computing(EC)is considered as an effective approach.However,due to the long transmission distance and resource-constrained computing services,the devices often suffer from the transmission delay and computation delay,which even causes different synchronization of tasks,thus it does not guarantee that the results in IoT applications can be back in real-time processing.Given the circumstance that some IoT devices generally are not fully utilized,this paper proposed a novel real-time scheduling edge resource pooling model(RSERP),i.e.,uploading real-time data stream to the resource-rich devices for collaborative processing based on the Device-to-Device(D2D)cooperation.In addition,in order to further optimize the cache queue of the real-time scheduling edge resource pooling model,this paper designed a real-time scheduling algorithm for adaptive data stream.According to real-time workload changes,the workload input to the cache is optimized by the task operation set.Rigorous theoretical analysis and extensive evaluations demonstrated the efficacy of the proposed solution.The main research contents and innovations are as follows:(1)For solving the resource allocation problem of IoT applications,this paper proposed a new real-time scheduling edge resource pooling model.According to the long-term resource occupancy of devices,the devices are divided into resource device and task device.In the task device,the cache is divided into two virtual queues.Part of the cached task is executed locally,and the other part of the cached task is transmitted to the resource device through D2 D offloading for execution,thus it can reduce the processing delay of the data stream.In addition,this paper formulated the long-term average sum energy consumption minimization problem mathematically under both computation and communication capacity constraints.(2)For solving the problem of long-term energy consumption optimization,this paper designed an online data stream scheduling algorithm based on Lyapunov optimization,which decomposed long-term problem into a series of NP-hard combination of optimization problems in each time slot.In order to solve this problems,this paper designed a greedy-based device matching algorithm.This paper conducted experiments under two types of network topology.Experiments show that the algorithm greatly reduces the sum energy consumption of the system while stabilizing the two virtual queues of task buffer.The experimental results demonstrate that the performance of the proposed algorithm is 8.6% better than the best result of random method and can approximate the exhaustive attack method under the different connection degree.(3)In order to further optimize the cache queue of the real-time scheduling edge resource pooling model,this paper designed a real-time scheduling algorithm for adaptive data stream.This paper proposed a joint optimization problem of minimizing the long-term average sum energy consumption and maximizing the long-term average sum utility,and obtained the optimal results by analytical method.The experiment shows that the algorithm reduces the longterm cache per device by 3.78%.