内容同步是分布式系统的基础性任务，已有方法通常采用sketch数据结构来压缩表示数据集合，在主机之间相互交换各自sketch之后鉴定并传输集合之间差异元素，从而实现各主机之间集合内容的一致性。然而，数据体量的指数增长以及数据类型的日益繁多使得已有同步方法无法在同步精度、时间开销和空间开销上满足各类分布式系统对内容同步的性能要求。为此，本课题开展以下研究：1）研究大规模动态集合成员的压缩表示问题，针对集合的不同特性，提出对应集合成员表示方法；2）研究针对多集的内容同步问题，将集合同步技术从简单集合扩展到多集领域；3）研究多方参与下的集合差分同步问题，形成多方同步的整体优势，最小化传输开销；4）研究关联数据集合差分同步问题，针对集合元素之间的不同关联特性，提出相应集合内容同步方法。通过以上研究为各类分布式系统实现精确、快速、高效的内容同步提供技术支撑，从而提高上层应用性能，保证服务质量。

Content synchronization is a fundamental task in distributed systems. The existing synchronization methods mainly represent the datasets with sketch data structures. After exchanging these sketches, the hosts identify and thereafter transfer the different elements among them, such that they reach the consistency upon their datasets. However, the data volume increases exponentially and the data types vary constantly, which makes the current synchronization methods incapable of meeting the requirements (including synchronization accuracy, time-consumption, and space overhead) put forward by current distributed systems. To this end, this project considers the following research problems: 1) the representation of large-scale dynamic datasets—propose differentiated representation scheme for diverse datasets; 2) the accurate synchronization of multisets—extend the synchronization problem from simple sets to multisets; 3) the content synchronization among multiple participants—design a synchronization method when there are more than two participants; 4) the synchronization of dependent datasets—investigate the customized synchronization methods according to the dependency between elements. With the above endeavor, this project will provide accurate, fast and efficient synchronization methods for distributed systems. As a consequence, the upper-level applications will achieve better performance and guarantee the user-level quality of service.