Abstract: To solve the problem of the recovery performance drop caused by data fragmentation in the deduplication-based cloud storage system,an adaptive fragmentation grouping was used to design a method for data recovery after deduplication in a cloud storage system.The approach was arranged not to use a fixed-size read window,to identify and delete the fragmentation,and to select a different reference chunk to minimize the number of containers.The number of referenced blocks could reduce disk access to accurately identify and remove fragmented data,improving data recovery performance and deduplication rates.The results show that,the approach can improve restore performance while reducing the amount of rewrite data.