对等云存储系统信誉机制研究

A reputation mechanism in peer-to-peer cloud storage systems

云存储是分布式存储技术与虚拟化技术结合的产物,是分布式存储技术的最新发展.云存储意味着存储可以作为一种服务,通过互联网提供给用户.相对于当前主从(Master/Slave)结构的GFS(Google File System),HDFS(Hadoop Distributed File System),Sector,KFS(Kosmos File System)等云存储系统,提出了一种基于DHT(Distributed Hash Table)的对等云存储系统MingCloud.重点研究了系统的信誉机制问题,具体包括综合信誉度的计算、信任模型设计、信誉机制的工作过程、安全性分析等方面.仿真实验表明,该机制具有较好的有效性和健壮性,能有效保护系统中的诚实节点,迅速降低恶意节点的信任值,抑制恶意交易行为.引入MingCredit之后,系统不再简单地以节点的物理性能诸如存储空间大小、网络带宽等方面的客观能力,以及在线时间作为评价标准,而是与节点参与到对等存储交易的行为特征密切关联起来.

As the latest development of the distributed storage technology,cloud storage is a product of the integration of distributed storage and virtualization technologies.Cloud storage allows use to storage facilities available on the internet.Relative to the current Master/Slave computing model cloud storage file systems,including GFS,HDFS,Sector and KFS,an open P2P（peer-to-peer） cloud storage system based on DHT was proposed,and we constructed a prototype system named MingCloud based on the Kademlia algorithm.According to the research-development needs of the system,a reputation mechanism was designed for the system,specifically including the calculation of reputation degree,trust model design,work process of reputation mechanism,and safety analysis.Our simulation results show that the mechanism of this model provides a high degree of effectiveness and robustness.Meanwhile,the reputation model can effectively protect honest nodes,reduce the trust value of malicious nodes,and restrain malicious behaviors.