Middle Map Architecture for vSAN Snapshot I/O Overhead

A middle map is introduced as an intermediary layer between the logical block address map and the physical block address map. The middle map stores mappings from logical block addresses to middle block addresses, while the physical map stores mappings from middle block addresses to physical block addresses. This intermediary structure allows the system to update only the middle map during segment cleaning operations, avoiding the need to update multiple snapshot logical maps simultaneously, thereby reducing I/O overhead while maintaining I/O performance.

The mapping architecture is segmented into three independent layers: the logical block address map, the middle map, and the physical block address map. Each layer can be updated independently. During segment cleaning, only the middle map needs to be updated to reflect new physical block assignments, while the logical block address map and snapshot logical maps remain unchanged. This segmentation allows snapshot accuracy to be maintained without increasing metadata write I/Os, as the middle map acts as a stable intermediary that decouples snapshot references from physical storage locations.

The mapping architecture transitions from a direct one-to-one mapping between logical and physical block addresses to a two-dimensional mapping structure involving middle block addresses. The logical block address map provides a stable interface for snapshot references, the middle map provides flexibility for physical block reassignment, and the physical map provides the actual storage locations. This dimensional expansion allows the system to maintain snapshot accuracy while enabling flexible segment cleaning operations without increasing overall device complexity.