Edge computing is becoming a major building block of next generation 5G/6G networks. However, infrastructure might not always be available because of slow deployment. At the same time, vehicular networks are becoming a reality now and cars are being equipped with a variety of short-range communication devices. The idea of vehicular micro clouds is to turn cars into (virtual) edge computing infrastructure. One of the challenging questions in this domain is to maintain data within and among such micro clouds. In this paper, we focus on this task and present a novel solution for such data exchange between vehicular micro clouds. For efficient operation, the dwell times of cars in such a micro cloud need to be known or accurately predicted. In an extensive study based on trace data, we investigate the distribution of dwell times of cars at intersections. We make use of this distribution as an input for designing an improved data exchange algorithm. As not all intersections are the same, adding additional variance further benefits the solution. We evaluated our algorithm in different vehicular densities, and we observed that we could maintain data 22–208% longer within the micro clouds using our new algorithm. Overall, our results show that our algorithm clearly outperforms previous solutions.

边缘计算正在成为下一代 5G/6G 网络的主要构建块。但是，由于部署缓慢，基础设施可能并不总是可用。与此同时，车载网络正在成为现实，汽车配备了各种短距离通信设备。车载微云的想法是将汽车变成（虚拟）边缘计算基础设施。该领域的一个具有挑战性的问题是在这些微云内部和之间维护数据。在本文中，我们专注于这项任务，并为车载微云之间的此类数据交换提出了一种新颖的解决方案。为了高效运行，需要知道或准确预测汽车在这种微云中的停留时间。在基于跟踪数据的广泛研究中，我们调查了汽车在交叉路口的停留时间分布。我们利用这种分布作为设计改进数据交换算法的输入。由于并非所有的交叉点都相同，因此添加额外的方差进一步有利于解决方案。我们在不同的车辆密度下评估了我们的算法，我们观察到我们可以使用我们的新算法在微云中将数据保存时间延长 22-208%。总的来说，我们的结果表明我们的算法明显优于以前的解决方案。我们观察到，使用我们的新算法，我们可以将微云中的数据维护时间延长 22-208%。总的来说，我们的结果表明我们的算法明显优于以前的解决方案。我们观察到，使用我们的新算法，我们可以将微云中的数据维护时间延长 22-208%。总的来说，我们的结果表明我们的算法明显优于以前的解决方案。