Privacy-aware Task Offloading And Resource Allocation For Cloud Edge End Computing

The big data processing problems caused by massive Internet of things devices can be efficiently solved under the cloud edge end architecture,yet this architecture has the characteristics of large-scale devices,heterogeneous resources,and complex hierarchy,which brings new challenges to privacy protection.Considering the collaboration between end devices,edge nodes,and remote clouds,the problem of task offloading and resource allocation in cloud edge end environment with privacy constraints is studied.It has important research significance and practical value.Challenges of the considered problems are listed as follows:(i)Since the resources of end devices are limited,some tasks need to be offloaded to the edge or cloud.However,under the constraints of the task and the bandwidth,how to make task offloading decisions reasonably to improve the quality of service is a difficult problem.(ii)The tasks offloaded to the edge server have different privacy requirements.How to formulate a reasonable resource allocation strategy to optimize the scheduling objective with the constraints of the task privacy is an NP-hard problem.For the problem of task offloading and resource allocation with privacy protection in the cloud edge end environment,a system model considering task privacy constraints is proposed.And the task offloading and resource allocation in different scenarios are studied.For independent tasks with no hard deadlines and privacy constraints,the problem of task offloading and resource allocation with the objective of minimizing the average completion time is considered.A privacy-aware heuristic algorithm of task offloading and resource allocation is proposed,which solves the offloading decision problem and resource allocation problem alternately and iteratively.It contains four components: task offloading sequence generation,offloading policy adjustment,communication resource allocation,computing resource allocation.Three offloading sequence generation rules are proposed based on the maximum local execution time,the maximum ratio of local execution time to transmission time,the minimum data size,respectively.On the basis of the load of edge servers,an offloading decision adjustment strategy is provided to control the number of low-privacy offloaded tasks.According to the total amount of uploaded data of each device,its required communication resources are estimated,and two task data transmission sequence generation rules are deployed: the minimum computation first and the minimum data size first.The local resource allocation and edge server resource allocation strategies are designed respectively.Matching methods for different privacy level tasks with edge server resources are also constructed.For independent tasks with hard deadlines and privacy constraints,the problem of task offloading and resource allocation with the objective of maximizing the number of successfully executed tasks is considered.The mathematical model of the problem is constructed.This problem can also be solved by the pre-described heuristics.Based on the characteristics of this problem,there are two offloading sequence generation rules: the latest deadline priority and the smallest latest start time priority.The earliest deadline priority rule is presented to get the order of task upload.In order to allocate the local resources reasonably,three rules are proposed that based on the minimum Computation,the earliest deadline,the minimum latest start time respectively.The priority queue is used to store tasks that need to be uploaded on each end device.And the first-come-first-served strategy is selected to schedule the task that arrive at the base station.The initial scheduling results are adjusted to maximizing the number of tasks which are executed successfully.The tasks that failed to execute are sorted by the latest start time of tasks in non-descending order,and the tasks are rescheduled on the local device in turn.To evaluate the performance of the proposed algorithm,the algorithm parameters are calibrated with Multi-factor ANOVA,and the optimal parameter combination is selected.The experimental results show that for the former problem,the proposed algorithm is better than the compared algorithms under different number of devices,task data volume and the proportion of different privacy levels of tasks.When the proportion of private tasks is relatively less,the proposed algorithm has outstanding performance.For the latter problem,the performance of the proposed algorithm is always better than other algorithms in many aspects,especially when the resources are sufficient.