移动业务特征认知及其分布模型:以即时消息为例

发布时间：2018-02-14 00:18

本文关键词： 业务特征即时消息统计规律建模重尾分布 α-稳定模型　出处：《中国科学:信息科学》2017年05期 　论文类型：期刊论文

【摘要】：众所周知,业务流量特点建模有助于设计能效低、稳定性强的网络协议.近年来新兴业务如雨后春笋一般涌现,重新考察这些新业务的特点也就至关重要了.本文以即时消息业务为例,了解当前移动互联网下的流量新特点.为了得到可信的结果,采集了大量的蜂窝网即时消息实测数据,并从用户层面和基站层面两个维度去研究其统计规律.首先在用户层面,通过用户级的数据包包长和到达间隔建模发现,相关结果同传统上包长符合几何分布、到达间隔满足指数分布的3GPP报告明显不同;与之相反,微信包长和到达间隔分别遵循幂律和对数正态分布.其次在基站层面,通过随机选取基站发现,α-稳定模型能更好地刻画基站层面基站流量——这一结果同传统固定宽带网络业务研究相吻合.最后,建立了用户层面数据包同基站层面流量的理论联系.
[Abstract]:As we all know, modeling the characteristics of service traffic is helpful in the design of network protocols with low energy efficiency and strong stability. In recent years, new services have sprung up. It is very important to re-examine the characteristics of these new services. This paper takes the instant messaging service as an example to understand the new characteristics of traffic under the current mobile Internet. In order to obtain credible results, a large number of real-time messaging data are collected in the cellular network. At the user level, through the modeling of packet length and arrival interval at the user level, it is found that the related results are consistent with the traditional geometric distribution of packet length. The 3GPP report of arrival interval satisfying exponential distribution is obviously different; on the contrary, the packet length and arrival interval of WeChat follow power law and logarithmic normal distribution respectively. Through random selection of base stations, it is found that the 伪 -stability model can better depict base station traffic at base station level, which is consistent with the traditional research on fixed broadband network services. Finally, the theoretical relationship between subscriber level data packets and base station traffic is established.
【作者单位】：浙江大学信息与电子工程学院;清华大学电子工程系;
【基金】：浙江省重点科技创新团队(批准号:2013TD20) 浙江省科技计划(批准号:2015C01075) 中国博士后创新人才支持计划(批准号:BX201600133)资助项目
【分类号】：TN929.5

【相似文献】