基于拍卖的边缘云期限感知任务卸载策略

Auction-based Edge Cloud Deadline-aware Task Offloading Strategy

随着万物互联和5G时代的到来,移动用户需要处理的数据量与其处理数据能力不匹配。将大量任务卸载到有限的边缘服务器上执行势必会产生竞争,拍卖模型的引入可以解决用户之间对资源的竞争问题。目前大多基于拍卖的任务卸载工作忽略了任务的期限感知,普遍的任务卸载工作只单一考虑延迟敏感任务,并且未考虑到保证卸载过程的安全性。基于此,提出了一种基于拍卖的期限感知任务卸载(Auction Based Deadline-aware Task Offloading,ABDTO)策略,利用基于智能合约的拍卖机制实现期限感知任务(延迟敏感型任务和非延迟敏感型任务)到边缘服务器的最优分配,以总效用(即总利润)作为评价标准,实现移动用户和边缘服务器的共赢。利用启发式遗传算法进行仿真实验,相比TACD,UPPER和RND算法,ABDTO策略的整体效用更高,最后利用Remix和Ganache等建立以太坊私有区块链网络进行仿真,证明了所提策略的正确性和可行性。

With the advent of the Internet of everything and the 5G era,the amount of data that mobile users need to process does not match their data processing capabilities,offloading a large number of tasks to limited edge servers for execution is bound to produce competition.The introduction of auction modelcan solve the problem of resource competition among users.At present,most task offloading works based on auction ignore the deadline perception of tasks,the general task offloading work only consi-ders delay-sensitive tasks,and does not consider ensuring the security of the offloading process.Therefore,an auction based deadline-aware task offloading(ABDTO)strategy is proposed,which uses the auction mechanism based on smart contract to realize the optimal allocation of deadline-aware tasks(delay-sensitive tasks and non-delay-sensitive tasks)to the edge ser-vers,and the total utility(i.e.total profit)is taken as the evaluationcriterion to achieve a win-win situation between mobile users and edge ser-vers.The heuristic genetic algorithm is used to conduct simulation experiments.Compared with TACD,UPPER and RND algorithms,ABDTO strategy has higher overall utility.Finally,using Remix,Ganache,etc.to establish the Ethereum private blockchain network for simulation,which proves the correctness and feasibility of the strategy.