In the last decade there has been an increasing interest and demand on the Internet of Things (IoT) and its applications. But, when a high level of computing and/or real time processing is required for these applications, different problems arise due to their requirements. In this context, low cost autonomous and distributed Small Board Computers (SBC) devices, with processing, storage capabilities and wireless communications can assist these IoT networks. Usually, these SBC devices run an operating system based on Linux. In this scenario, container-based technologies and fog computing are an interesting approach and both have led to a new paradigm in how devices cooperate, improving overall capacity in a cluster of these SBC devices. The use of containers is considered a lightweight virtualization, allowing an application to be broken into small tasks as services, enabling load balancing, flexibility and scalability. Nevertheless when the number of devices and containers increases in the cluster, it is required an orchestration layer. There are not many solutions and available alternatives using these technologies applied on these networks, and less an assessment of their performances. This paper focuses on these technologies when we use fog computing with low cost SBC devices in a context of IoT. We use Linux containers and different available orchestration platforms (in particular Docker Swarm and Kubernetes), to run on the top of the cluster of commercial SBC devices. Thus, we carry out a thorough functional and performance comparison with different real topologies (wired and wireless) and using both homogeneous and heterogeneous clusters of SBC devices, showing their results. We conclude that with the collected experimental results, Docker Swarm orchestration platform outperforms its counterparts in the scenarios shown.

在过去的十年中，人们对物联网（IoT）及其应用的兴趣和需求不断增长。但是，当这些应用需要高水平的计算和/或实时处理时，由于它们的要求而产生了不同的问题。在这种情况下，具有处理，存储功能和无线通信的低成本自主和分布式小型板计算机（SBC）设备可以帮助这些IoT网络。通常，这些SBC设备运行基于Linux的操作系统。在这种情况下，基于容器的技术和雾计算是一种有趣的方法，并且两者都导致了设备如何协作的新范例，从而提高了这些SBC设备群集中的整体容量。容器的使用被认为是轻量级的虚拟化，允许将应用程序分解为服务的小任务，从而实现负载平衡，灵活性和可伸缩性。但是，当集群中的设备和容器的数量增加时，则需要一个业务流程层。在这些网络上应用这些技术的解决方案和可用替代方案并不多，其性能评估也较少。当我们在物联网环境中将雾计算与低成本SBC设备结合使用时，本文将重点介绍这些技术。我们使用Linux容器和其他可用的编排平台（特别是Docker Swarm和Kubernetes）在商业SBC设备集群的顶部运行。从而，我们对不同的实际拓扑（有线和无线）进行了彻底的功能和性能比较，并使用了SBC设备的同质和异类群集，以显示其结果。我们得出的结论是，在收集的实验结果中，Docker Swarm编排平台在所示方案中优于其同类产品。