Big Data Synchronization among Isolated Data Servers in Disaster
Big Data Synchronization among Isolated Data Servers in Disaster
复制标题
DOI：
10.1109/glocom.2017.8254994
复制
发表时间：
2017-12
期刊：
影响因子：
--
通讯作者：
Kazuya Anazawa;T. Miyazaki;Peng Li;Xiaoyan Wang
中科院分区：
文献类型：
--
作者：
Kazuya Anazawa;T. Miyazaki;Peng Li;Xiaoyan Wang
当大规模灾难发生时，由于现有网络基础设施的严重破坏，有效的网络连接和通信变得困难。与此同时，在这样的灾害环境中，人们对疏散和救灾活动中的信息共享有着强烈的需求。为了满足这些繁重的通信需求，建立由便携式服务器组成的局域网（LAN）被认为是最有前途的解决方案之一。基于已建立的局域网，人们可以在覆盖区域内共享灾害相关信息。但由于缺乏稳定的互联网连接，这些局域网是孤立的，不能真实的时间同步。为了解决这个问题，在本文中，我们提出了一个间歇性的数据同步方案，通过引入移动车辆作为中继，在灾难发生后，孤立的数据服务器之间交换数据。以最大化移动的中继能力约束下的同步加权数据量为目标，我们制定了一个随机规划问题的轨迹规划。我们利用混沌理论和李雅普诺夫漂移技术来解决这个问题，在一个在线的设置，这是实用的真实的灾难环境。我们的理论分析表明，我们提出的在线算法的性能差距是O（1/V）$的最佳。此外，大量的模拟和与其他算法进行比较，以显示我们提出的在线算法的上级性能。
When a large-scale disaster happens, efficient network connection and communication becomes difficult due to serious damage of existing network infrastructures. Meantime, people have strong demands of information sharing with each other for evacuation and disaster-relief activities in such a disaster environment. To serve these heavy communication demands, establishing local area networks (LANs) consisting of portable servers has been considered as one of the most promising solutions. Based on the established LANs, people can share disaster-related information in covered area. However, due to the lack of stable Internet connection, these LANs are isolated and cannot be synchronized in real time. To tackle this problem, in this paper, we propose an intermittent data synchronization scheme by introducing moving vehicles as relays to exchange data between isolated data servers after disasters. With the objective of maximizing the synchronized weighted data volume under the capability constraints of the mobile relay, we formulate a stochastic programming problem for trajectory planning. We leverage queueing theory and the Lyapunov-drift technique to solve this problem in an online setting, which is practical for a real disaster environment. Our theoretical analysis shows that the performance gap of our proposed online algorithm is $O(1/V)$ of the optimum. Additionally, extensive simulations and comparisons with other algorithms are conducted to show the superior performance of our proposed online algorithm.