基于服务更新的异构任务卸载方法  

Heterogeneous task offloading method based on service update

作　　者：韩超臣 张俊娜[1] 王欣新 袁培燕[1] 赵晓焱[1] 刘春红[1] Han Chaochen;Zhang Junna;Wang Xinxin;Yuan Peiyan;Zhao Xiaoyan;Liu Chunhong(College of Computer&Information Engineering,Henan Normal University,Xinxiang Henan 453007,China)

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007

出　　处：《计算机应用研究》2024年第8期2481-2488,共8页Application Research of Computers

基　　金：国家自然科学基金资助项目(62072159);河南省科技攻关项目(232102211061,222102210011)。

摘　　要：边缘计算将存储和计算资源下沉到网络边缘,用户可将时延敏感型和计算密集型应用程序的任务卸载到边缘服务器执行,从而降低时延和能耗。已有的任务卸载研究通常忽略卸载任务的异构性,且默认边缘服务器缓存的服务能够长期满足用户的服务需求。然而,不同的任务需要不同的服务提供执行环境,且边缘服务器资源受限只能部署少量服务。因此,为了最小化时延和能耗(即成本),提出了一种联合服务更新和云边端协作的异构任务卸载方法。首先,通过改进的页面置换算法,预测出潜在的用户服务需求量,及时更新边缘服务器的服务。其次,在云边端协作基础上,通过改进的贪婪算法完成任务卸载,且使得成本最小。最后,基于真实数据集进行了充分的实验,实验结果表明,与对比方法相比,所提方法能够降低成本7%~16%。Edge computing brings storage and compute resources down to the edge of the network.Users can offload latency-sensitive and compute-intensive applications to the edge server for execution,thereby reducing latency and energy consumption.Existing task offload studies often fall short.The first point is that the existing research on task offloading often overlooks the heterogeneity of offloading tasks.The second point is that the default edge server caching service can meet the long-term service needs of users.However,different tasks require different services to provide execution environments.And the finite resources of edge servers can only deploy a small number of services.Therefore,to minimize latency and energy consumption(i.e.cost),this article proposed a heterogeneous task offloading method that combined service updates and cloud edge collaboration.The method consisted two following main parts.The first part was an improved page replacement algorithm.The algorithm could predict the potential user service demand and updated the service of the edge server in time.The second part was the collaboration between the cloud-side ends,offloading tasks through collaboration and keeping minimum total costs.Finally,it conducted thorough experiments using real datasets.The experimental results show that the proposed method can reduce the total cost by 7%to 16%compared to the comparison methods.

关 键 词：边缘计算 云边端协作 服务更新 改进的页面置换算法 改进的贪婪算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]