Data reduction combines inline compression, inline deduplication, pattern detection, efficient data placement, and machine learning (ML). This feature combination allows users to write host data amounts that exceed the total physical capacity available while achieving the performance expected from an enterprise storage system. This feature is enabled by default, and you can disable or reenable it at the storage group level. The data reduction feature supports all data services available in PowerMax 2500 and 8500 systems. This support also applies to count-key-data (CKD) emulation, but it does not include deduplication for CKD.

Compression reduces the size of data, while deduplication stores data as a single instance. Pattern detection includes a nonzero allocate function that excludes strings of consecutive zeros from being stored as part of the compressed data. Compression, deduplication, and pattern detection are performed using hardware assistance integrated within the system to reduce the overhead of performing these functions. ML identifies the busiest data stored on disk and ensures that it remains unreduced for optimal performance. Efficient data placement uses a function called compaction, which strategically stores data to minimize wasted space and reduces the need for garbage collection or defragmentation functions.