针对传统连续数据保护技术中使用的快照存储技术难以高效存储足够多快照的问题,提出了一种新的快照存储技术,即基于凸点集合的快照存储技术(convex point SNAPshot,简称CSNAP).在讨论了CSNAP 技术使用的数据结构,并引进了在数据结构中加入指针构成的凸点概念的基础上,分析了凸点集合具有的基本性质,并给出了利用凸点集合储存快照的相关算法,同时还给出了一种使用逆行代价改进CSNAP 的方法.最后通过在实际工作负载和随机合成工作负载下的模拟实验说明了,在平均情况下,使用改进的CSNAP 技术,可以将快照占用的储存空间减少到传统方法的10%以下.

This paper proposes a new snapshot method for continuous data protection (CDP) system that considers the disability of taking large amount of snapshots in traditional CDP systems. The snapshot method (Convex Point SNAPshot, CSNAP) is based on the concept of convex point set. After the data structure of CSNAP and introduced the concept of convex point based on the pointers in the data structure have been discussed, the study analyzes the properties of convex point set and proposed CSNAP algorithms. An enhanced CSNAP method by introducing the concept of retro-cost is also proposed. Finally, the study uses a typical workload and random generated trace data to test CSNAP method. The experimental results show that at average CSNAP takes less than 10% storage space of traditional snapshot method.