Multiobjective Optimization for Computation Offloading in Fog Computing
Multiobjective Optimization for Computation Offloading in Fog Computing
复制标题
DOI：
10.1109/jiot.2017.2780236
复制
发表时间：
2018-02-01
影响因子：
10.6
通讯作者：
Ristaniemi, Tapani
中科院分区：
文献类型：
--
作者：
Liu, Liqing;Chang, Zheng;Ristaniemi, Tapani
雾计算系统是为实现物联网提供计算、存储、控制和联网能力的新兴体系结构。在雾计算系统中，移动的设备（MD）可以将其数据或计算昂贵的任务卸载到其附近的雾节点，而不是远程云。虽然卸载可以减少MD处的能量消耗，但是它也可能招致更大的执行延迟，包括MD与雾/云服务器之间的传输时间以及服务器处的等待和执行时间。因此，如何平衡网络的能量消耗和时延性能具有重要的研究意义。此外，基于能量消耗和延迟，如何设计一个成本模型的MD享受雾和云服务也很重要。本文利用排队论对雾计算系统中卸载过程的能耗、执行延迟和支付代价进行了深入研究。具体而言，三个排队模型，分别适用于MD，雾，云中心，并明确考虑无线链路的数据速率和功耗。在理论分析的基础上，提出了一个多目标优化问题，该问题以最小化能量消耗、执行延迟和支付成本为联合目标，通过为每个MD寻找最优卸载概率和发射功率来实现。大量的仿真研究表明，该方案的有效性和上级的性能比几个现有的计划进行了观察。
Fog computing system is an emergent architecture for providing computing, storage, control, and networking capabilities for realizing Internet of Things. In the fog computing system, the mobile devices (MDs) can offload its data or computational expensive tasks to the fog node within its proximity, instead of distant cloud. Although offloading can reduce energy consumption at the MDs, it may also incur a larger execution delay including transmission time between the MDs and the fog/cloud servers, and waiting and execution time at the servers. Therefore, how to balance the energy consumption and delay performance is of research importance. Moreover, based on the energy consumption and delay, how to design a cost model for the MDs to enjoy the fog and cloud services is also important. In this paper, we utilize queuing theory to bring a thorough study on the energy consumption, execution delay, and payment cost of offloading processes in a fog computing system. Specifically, three queuing models are applied, respectively, to the MD, fog, and cloud centers, and the data rate and power consumption of the wireless link are explicitly considered. Based on the theoretical analysis, a multiobjective optimization problem is formulated with a joint objective to minimize the energy consumption, execution delay, and payment cost by finding the optimal offloading probability and transmit power for each MD. Extensive simulation studies are conducted to demonstrate the effectiveness of the proposed scheme and the superior performance over several existed schemes are observed.