The core innovation of SnapFiles is the decoupling of data from metadata. When a client writes a file, the data stream is processed by a chunking algorithm (using variable-sized content-defined chunking to handle shifting data boundaries).