The present invention provides a data deduplication method, apparatus and system. The method comprises: carrying out coarse-grained data block division on a target file, so as to divide the target file into a plurality of coarse-grained data blocks; carrying out duplicate data block detection on the coarse-grained data blocks, and obtaining a first result; based on the first result, carrying out fine-grained data block division on each non-repetitive coarse-grained data block in the plurality of coarse-grained data blocks, so as to divide the non-repetitive coarse-grained data block into a plurality of fine-grained data blocks; carrying out duplicate data block detection on the fine-grained data blocks, and obtaining a second result; and based on the second result, storing non-repetitive fine-grained data blocks in the plurality of fine-grained data blocks, wherein the size of each coarse-grained data block is greater than the size of each fine-grained data block, and the demarcation of the coarse-grained data block division belongs to the demarcation of the fine-grained data block division.