Abstract: Concerning the scenario of localized geographical area, the autonomous mobile cloud is capable of providing provisional storage services for cloud users. However, because of the limitation of storage space in mobile nodes, maintaining the data persistence becomes more difficult. This paper proposes a space-limited storage protocol for  autonomous mobile cloud named as SLAMS protocol. The central idea of this protocol is that K copies for each data block are kept via periodic communications among mobile nodes, and data block copies are stored into nodes with larger free storage space. This paper presents the specific data copy maintaining mechanism, corresponding data structures and algorithms, and detailed design of SLAMS protocol, respectively. The simulation results show that, compared with current major solution Phoenix, the data lost rate can be reduced by up to 50%, the time of convergence can be reduced by up to 95%, and the overhead of data block storage can be reduced by up to 90% in the SLAMS protocol. Besides, the SLAMS protocol can effectively increase the probability of K copies by up to 40% for each data block in case that mobile nodes have limited storage space.

Key words: autonomous mobile cloud, storage service, data copy maintenance, data loss rate

摘要： 自治移动云可以为本地区域内的移动用户提供临时的存储服务。在节点空间有限的情况下，保持数据的持久性是自治移动云存储服务的一个主要挑战。提出了一种有限空间下的自治移动云存储协议，称为SLAMS（space limited storage protocol for autonomous mobile cloud），其基本思想是通过移动节点之间周期性的信息交互，维护每个数据块在网络中的K个副本，将数据块副本按需存储在空闲空间较大的节点中。给出了SLAMS协议中的副本维护机制、数据结构和算法、SLAMS协议的详细设计。仿真实验结果表明，相对于现有的自治移动云存储协议Phoenix，SLAMS协议在节点空间有限的情况下，具有较低的数据丢失率，最大可降低50%左右；具有较短的收敛时间，最高可降低95%；具有较少的数据块存储开销，最高可降低约90%；而且SLAMS协议中数据块维持[K]个副本的概率明显提高，最大可提高40%左右。

关键词: 自治移动云, 存储服务, 数据副本维护, 数据丢失率