数据海量化是信息技术发展的必然结果，分布式云存储已成为海量数据存储的有效手段。然而，云存储系统中多个节点同时故障的情况时有发生，故障节点的数据丢失将给用户带来无法估量的损失，制约分布式云存储系统进一步发展。作为提高云存储系统可靠性的关键技术，局部再生码能有效修复存储系统中的节点故障，同时降低修复的磁盘I/O开销，但其局部码修复多个节点故障时带宽和时间开销大，难以满足云存储系统在故障修复速度和带宽方面需求。针对该问题，首先构造能同时高效修复多个故障节点的局部码，同时给出高容错性局部再生码的设计方法，建立局部再生码的云存储容错架构；其次，设计基于高容错性局部再生码的高效故障修复算法，实现多个云存储节点故障修复；最后，为适应实际分布式云存储系统异构性的特点，研究高容错性局部再生码的可扩展性构造方法。本课题将进一步完善分布式云存储系统中数据可靠存储理论，为分布式云存储的实用化和产业化提供技术支撑。

As massive data becomes an inevitable development result of information technology, distributed cloud storage has proven to be the most effective means of massive data storage. However, since multiple node failures occur frequently in distributed cloud storage system, it will restrict the development of distributed cloud storage that the loss of cloud storage data will bring immeasurable damage to users. As a key technology to improve the reliability of cloud storage system, local regenerating codes can repair the node failures of distributed storage system effectively, while reducing disk I/O overhead. But when their local codes repair multiple node failures, the higher bandwidth and time overhead cannot meet the repair requirements of distributed cloud storage system in the speed and bandwidth of failure repair. In order to solve this problem, local regenerating codes with high fault tolerance are proposed firstly, in which their local codes can repair multiple node failures simultaneously, and cloud storage architecture is constructed based on local regenerating codes. Moreover, fault repair algorithm based on local regenerating codes is designed, to repair multiple node failures of cloud storage system efficiently. Finally, to adapt to the characteristics of the heterogeneity of actual distributed cloud storage system, the method of extensible construction for local regenerating codes with high fault tolerance is studied.. This project will further improve the data reliable storage theory in distributed cloud storage system, and provide technical support for the practicality and industrialization of distributed cloud storage.