Time-varying resource allocation in which the resource requirements of a job can vary over time is a new challenge in cloud computing. Time-varying resource allocation can be combined with an auction mechanism to improve the social welfare of resource providers. However, existing research results are based on fixed resource requirements and consequently cannot be used in time-varying resource allocation. This paper proposes a creative integer programming model for time-varying resource allocation problems and designs a strategy-proof online auction mechanism that allows jobs to be scheduled in a preemptive-restart mode. The advantage of this approach is that it can respond to high-priority jobs in a timely manner while still executing low-priority jobs with the restart mode. For the resource allocation and scheduling algorithm, we propose dynamic priority based on the dominant resource proportion and valid active time to improve social welfare and resource utilization. Furthermore, we present a payment pricing algorithm based on critical value theory. Finally, we prove that our proposed mechanism is strategy-proof. Our approach is experimentally compared with existing algorithms in terms of execution time, social welfare, resource utilization and job completion ratio.

随时间变化的资源分配，其中作业的资源需求会随着时间的推移而变化，这是云计算中的一个新挑战。时变资源分配可以与拍卖机制相结合，以提高资源提供者的社会福利。然而，现有的研究成果是基于固定的资源需求，因此不能用于时变资源分配。本文针对时变资源分配问题提出了一种创造性的整数规划模型，并设计了一种策略证明在线拍卖机制，允许以抢占式重启模式调度作业。这种方式的优点是可以及时响应高优先级的作业，同时仍然以重启模式执行低优先级的作业。对于资源分配和调度算法，我们根据优势资源比例和有效活跃时间提出动态优先级，以提高社会福利和资源利用率。此外，我们提出了一种基于临界值理论的支付定价算法。最后，我们证明了我们提出的机制是策略证明的。我们的方法在执行时间、社会福利、资源利用率和作业完成率方面与现有算法进行了实验比较。