在大数据背景下，非结构化云数据管理系统中数据节点需要处理不断膨胀的原 始数据、索引数据和中间数据，“数据膨胀”将显著增加云数据管理系统的时间和能耗等 各类开销． 为降低非结构化数据管理系统因数据频繁移动而导致的数据传输开销，文中提 出了一种不稳定数据分区的识别算法． 首先面向非结构化数据管理系统，通过引入云模型 理论对存储系统中的数据分区进行云建模，识别出不稳定的数据分区，然后调用相关算法 对其进行重新布局． 实验结果显示，不稳定数据分区识别算法可以有效地识别出不稳定的 数据分区，对其重新布局后，降低数据传输开销的效果显著．

In the context of big data，data nodes in unstructured cloud data management systems need to process ever-expanding raw data，index data and intermediate data． “Data expansion”will significantly increase the time and energy consumption of cloud data management systems． In order to reduce the data transmission overhead caused by the frequent movement of data in unstructured data management system，an algorithm for identifying unstable data partitions was proposed． Firstly，for the unstructured data management system，the cloud model of data partition was conducted in the storage system by introducing the cloud model theory，and unstable data partitions were identified． Then the relevant algorithm was called to re-layout unstable data partitions． Experimental results show that the unstable data partition identification algorithm can effectively identify unstable data partitions and relayout them，and significantly reduce data transmission overhead．