Recently, the issue of offloading cellular data while reducing the duplicated cellular transmission has gained more and more attention. Several studies have shown that sharing contents through Device-to-Device (D2D) to offload traffic to local connections nearby can offer better performance for mobile users. Nevertheless, most existing proposals are somewhat confined to small-scale data sets or limited feature dimensions, relied on unconsolidated hypotheses and measurements of data sets. This paper presents a prior work of large-scale measurements on 3.56 TBytes of real-world data sets, which contain D2D content sharing activities from a popular D2D sharing application (APP). We conduct a comprehensive analysis of multi-dimensional features, including time series, structural properties, meeting dynamics, location relationship, and propagation tree. Our analysis reveals that (i) D2D sharing makes the hops between users shorter (in 3 or 4 degrees), (ii) the maximum spreading distance of content dissemination is 27 hops, (iii) we provide a new evidence of log-normal distribution of all user encounters (named meeting dynamics in this paper) based the fit of inter-TX time, Inter-Content Time (ICT) and Contact Time, (iv) online factor (O) and social factor (S) demonstrate the largest positive correlation and indicate that the two factors have high linear correlation. Finally, we analyze the correlations among all the impact factors by Pearson coefficient, principal component analysis, and latent semantic analysis, respectively. Results reveal that online factor (O) and social factor (S) have a high correlation, especially both of them have a great effect on D2D sharing activities.

中文翻译：

---

基于移动社交网络中设备间共享的大规模离线移动应用传播的测量与分析

近来，在减少重复的蜂窝传输的同时卸载蜂窝数据的问题越来越受到关注。多项研究表明，通过设备到设备（D2D）共享内容以将流量分流到附近的本地连接可以为移动用户提供更好的性能。但是，大多数现有建议都依赖于未合并的假设和数据集的度量，仅局限于小规模的数据集或有限的特征尺寸。本文介绍了对3.56 TB实际数据集进行大规模测量的先验工作，这些数据集包含来自流行的D2D共享应用程序（APP）的D2D内容共享活动。我们对多维特征进行了全面的分析，包括时间序列，结构特性，会议动态，位置关系和传播树。i）D2D共享使用户之间的跃点更短（3或4度），（i i）内容分发的最大传播距离为27跃点，（i i i）我们提供了所有对数正态分布的新证据根据TX间时间，内容间时间（ICT）和联系时间（i v）在线因素（O）和社会因素（S）表现出最大的正相关，并表明这两个因素具有较高的线性相关性。最后，我们分别通过皮尔逊系数，主成分分析和潜在语义分析来分析所有影响因素之间的相关性。结果表明，在线因素（O）和社交因素（S）具有高度相关性，尤其是二者对D2D共享活动的影响很大。