Abstract: In the traditional blocklevel continuous data protection (CDP) system, once the recovery task at a certain point in time has been completed, the data state of the protected disk at the current time is completely covered. The current work can continue which means that the protected disk must be restored to the latest state again. Aiming at the problem of “oneview, tworecovery” of the traditional CDP system recovery function, a log disk partitioning technology and a metadata hierarchical multilevel indexing technology are proposed. The metadata and the changed data are respectively stored in different areas of the log disk, and the metadata block and the variable data block are organized by using the B+ tree and the improved B+ tree to form a hierarchical multilevel index of the metadata, and any point in the past will be used. The data is mounted to a virtual disk for user viewing. It has been proved by experiments that the method provides users with data at any point in time without affecting the current data state of the protected disk, which simplifies operations of users and reduces their time consumption.

Key words: continuous data protection (CDP), log disk partition, metadata, B tree, index, virtual disk

摘要： 在传统的块级持续数据保护(continuous data protection, CDP)系统中，过去某时间点的恢复任务一旦完成，被保护磁盘当前时刻的数据状态就被完全覆盖，必须将被保护磁盘再次恢复到最新数据状态，当前工作才能得以继续进行下去.针对传统CDP系统恢复功能“一次查看，两次恢复”的问题，提出了日志盘分区技术及一种元数据层次式多级索引技术.将元数据与变动数据分别存放在日志盘的不同区域，并利用B+树与改进B+树将元数据区块和变动数据区块组织起来，形成元数据层次式多级索引，将过去任意时间点的数据挂载到虚拟磁盘供用户检视.经实验证明，该方法在不影响被保护磁盘当前数据状态的前提下向用户提供过去任意时间点的数据，简化了用户操作，减少了时间消耗.

关键词: 持续数据保护, 日志盘分区, 元数据, B 树, 索引, 虚拟磁盘