To satisfy the increasingly complex demands of the Internet of Things (IoT) applications, multiclouds are a promising solution that can provide scalable, various, and abundant resources. However, in multiclouds, each cloud has its specific virtual machine (VM) type and pricing scheme. In addition, the request arrival, network bandwidth, and VM’s price all vary with time and are hardly predicted. In such cases, the request scheduling and resource provisioning (RSRP) for cost efficiency becomes a highly challenging work. In this article, to capture the dynamics in the multiclouds environment, we formulate a stochastic optimization problem where the aim is to minimize the system cost and guarantee the IoT applications’ queueing delay. By applying stochastic optimization theory, the original problem is transformed into a deterministic optimization problem in each slot, and then the deterministic problem is further decomposed into three independent subproblems. An online RSRP algorithm is devised to obtain these subproblems’ optimal solutions. Mathematical analysis shows that RSRP can approach the optimal system cost while bounding the queueing delay, and make an arbitrary tradeoff between system cost and queueing delay as well. Moreover, trace-driven simulation results show the effectiveness of RRSP.

中文翻译：

---

物联网多云中的经济高效的请求调度和资源配置

为了满足物联网（IoT）应用程序日益复杂的需求，多云是一种有前途的解决方案，可以提供可扩展的，多样的和丰富的资源。但是，在多云中，每个云都有其特定的虚拟机（VM）类型和定价方案。此外，请求的到达，网络带宽和VM的价格都随时间变化，并且很难预测。在这种情况下，提高成本效率的请求调度和资源供应（RSRP）成为一项极富挑战性的工作。在本文中，为了捕获多云环境中的动态，我们提出了一个随机优化问题，其目的是最小化系统成本并确保IoT应用程序的排队延迟。通过应用随机优化理论，将原始问题转换为每个时隙中的确定性优化问题，然后将确定性问题进一步分解为三个独立的子问题。设计了一种在线RSRP算法来获得这些子问题的最佳解决方案。数学分析表明，RSRP可以在限制排队延迟的同时逼近最佳系统成本，并且可以在系统成本和排队延迟之间做出任意权衡。此外，跟踪驱动的仿真结果显示了RRSP的有效性。并在系统成本和排队延迟之间做出任意权衡。此外，跟踪驱动的仿真结果显示了RRSP的有效性。并在系统成本和排队延迟之间做出任意权衡。此外，跟踪驱动的仿真结果显示了RRSP的有效性。