Finding entities of interest in indoor commercial places, such as the merchandise in supermarkets and shopping malls, is an essential issue for customers, especially when they are unfamiliar with an ad hoc indoor environment. This type of location-based indoor service requires comprehensive knowledge of indoor entities, including locations as well as their semantic information. However, the existing indoor localization approaches fail to directly localize these general entities without dedicated devices. This paper first focuses on the problem of discovering large-scale general entities of interest in indoor commercial spaces without pre-deployed infrastructure. We present a unique entity localization approach that leverages the localization results from multiple independent users to accurately determine the location of corresponding entities. Our key idea is to exploit the short-distance estimation with dead reckoning to guarantee the accuracy of entity localization. We develop a prototype system based on the crowdsourcing method, iScan, and test it in one of the biggest supermarkets in Changsha, China, to validate the performance of our design. Extensive experimental results show that our approach can achieve meter-level accuracy in a single day with 70 participants. Moreover, in a monthly evaluation with 500 effective participants, iScan discovered more than 200 entities and localized approximately 75% of them within 2 m.
作为云计算的基础设施和下一代网络技术的创新平台,数据中心网络的研究成为了近年来学术界和工业界关注的热点.文中围绕数据中心网络研究的基本问题,介绍了国际国内的研究现状,包括数据中心网络拓扑设计、传输协议、无线通信、增强以太网、虚拟化、节能机制和软件定义网络(Software Defined Networking,SDN)等,并展望了数据中心网络的发展趋势.
由于当前数据中心网络应用对延迟要求很高,但是数据中心网络采用TCP作为其传输协议,从而引发了延迟敏感流的长尾效应问题,故提出一种基于编码的传输协议(code-based transport protocol,CTP).该协议支持选择性反馈,旨在避免过大的RTO值所导致的长尾效应,同时减少数据中心网络间的传输延迟.CTP方案采用UDP替代TCP作为其传输协议,提出对数级别反馈的LT编码(logarithmic-feedback LT code,LFLT编码)保证可靠性,同时设计了一个有效的自适应的拥塞控制算法提升传输效率并且可以和TCP流友好共存.实验结果表明,CTP在编码效率上比传统LT编码提升了100%,传输时间方面比TCP和基于纯LT编码的UDP分别提升了150%和50%,受丢包率的影响小很多,更适用于流量波动较大的数据中心网络.
