Osmotic Computing represents a glue solution able to manage the deployment and orchestration of interconnected microelements across heterogeneous physical and virtual infrastructures (e.g., IoT, Edge and Cloud nodes) according to the behavior of hardware and software components during the time. The adoption of Osmotic Computing is challenging, but addressing networking issues is a key research topic due to the emergence of new problems in terms of QoS requirements. In this paper, we analyze how to exploit well-known networking solutions, such as the Dijkstra’s algorithm, and Big Data oriented technologies, such as the Hadoop and MapReduce, to provide efficient newtorking functionalities in Osmotic Computing. In particular, our objective is to minimize the routing path computation time in the software defined network (SDN) at the basis of microelement networking, as well as to ensure a global view and a high level of dynamism of our network topology. To accomplish this task, we process routing tables through a MapReduce based implementation of the Dijkstra’s algorithm whenever a topology change occurs, and we export routing results into the SDN. Our experimental results show that our networking strategy drastically reduces the best path computation time whenever the network of microelements is very large.

渗透计算代表一种粘合解决方案，能够根据当时的硬件和软件组件的行为来管理跨异构物理和虚拟基础架构（例如IoT，Edge和Cloud节点）的互连微元素的部署和编排。渗透计算的采用具有挑战性，但是由于QoS要求方面的新问题的出现，解决网络问题是一个关键的研究主题。在本文中，我们分析了如何利用著名的网络解决方案（例如Dijkstra的算法）以及面向大数据的技术（例如Hadoop和MapReduce）来在渗透计算中提供有效的更新功能。特别是，我们的目标是在微元素网络的基础上最大程度地减少软件定义网络（SDN）中的路由路径计算时间，并确保对我们的网络拓扑结构具有全局性和高度的动态性。为了完成此任务，只要拓扑发生变化，我们就通过基于MapReduce的Dijkstra算法的实现来处理路由表，并将路由结果导出到SDN中。我们的实验结果表明，只要微元素的网络很大，我们的网络策略就会极大地减少最佳路径计算时间。然后将路由结果导出到SDN。我们的实验结果表明，只要微元素的网络很大，我们的网络策略就会极大地减少最佳路径计算时间。然后将路由结果导出到SDN。我们的实验结果表明，只要微元素的网络很大，我们的网络策略就会极大地减少最佳路径计算时间。