【作者】 卢冰；

【导师】 吴松；

【作者基本信息】 华中科技大学 ， 计算机系统结构， 2012， 硕士

【摘要】 车联网是物联网领域的一个典型应用，其主要基于SOA架构的应用定制服务。涉车信息资源是国家战略资源，其应用将对我国国民经济活动和社会管理服务产生重大影响，对构建中国“物联网”，发挥基础性支撑作用。在车联网系统中，如何对涉车数据进行管理是其中一个必须解决的关键问题。车联网数据存储及复杂查询系统作为车联网后台的数据处理子系统，管理整个涉车信息资源，为上层云平台的各种应用提供查询服务和为数据挖掘分析提供支持。系统的主要技术关键点在于对海量历史数据进行有效管理的同时，能够针对应用的多样性提供简单可靠的查询接口。车联网数据存储及复杂查询系统基于著名的NoSQL分布式存储系统Cassandra，并把车联网的原始数据根据其特点划分为静态数据和动态数据，对Cassandra的数据存储策略进行了改进：对于动态数据，采用了分布式存储；对于静态数据，则采用了完全本地化存储，即每个节点都保留一份。在此基础上，系统还针对车联网应用的特点增加了一些复杂查询的接口，支持对数据的连接查询，尽可能减少上层应用和存储系统的交互次数，降低了应用服务器的负载和网络传输的开销。对系统的测试表明，系统对静态数据的查询速度有了显著的提升，同时，上层应用使用复杂查询接口的数据访问速度也比仅仅采用原型系统Cassandra原有接口时有了近30%的效率提升。更多还原

【Abstract】 Vehicle-Internet is a typical application in Internet Of Things. It customizes servicesto the public through a SOA-based architecture. Vehicle related information is a nationalstrategic resource and its application will have a significant impact on China’s nationaleconomic activity and social services, and will play a fundamental role in building China’sInternet of Things. In Vehicle-Internet, how to manage these data is one of the key issuesthat must be addressed. As a subsystem of background data processing, Vehicle-Internetdata storing and complex queries system manage the whole car-related informationresources, provide support for all sorts of applications in upper cloud platforms, includingquerying and data mining. The main technical point is on one hand effectively managingoceans of historical data, and on the other hand provide simple and reliable queryinginterface to various applications.Vehicle-Internet data storage and complex queries system is based on a famousNoSQL style distributed storage system ‘Cassandra’, and it divided the raw data inVehicle-Internet into two parts: dynamic data and static data. It also improves its datastoring policies and upgrades some data locality: adopting distributed storage method forthe dynamic data while using a fully localized storage strategy for the static data.Furthermore, the system provides some complex query interfaces based on the features ofthe applications to support the join operation among the raw data, thereby lowering theserver’s load and the network transmission overhead. It also improves its data storingpolicies and upgrade some data locality. This system adopts different storing techniques todifferent raw data in Vehicle-Internet, so it could provide simple, uniform complexquerying interface to system users while lowering server’s load and increasing querying’sspeed.The results of the system evaluation demonstrate that our method significantlyimproves the query performance of static data. At the meantime, the up-layer applicationsusing new complex query API outperform the ones that using native interface of cassandraby30%on average.更多还原