All I/O is passed through cache and then processed by the system. Data reduction actions are performed after the data is received by the system and before it is placed on disk. The inline process requires additional checks within the I/O flow where data reduction applies. The system uses these checks to determine whether incoming data needs to pass through the data reduction hardware. Incoming data for a storage group with data reduction enabled will follow the data reduction flow. However, due to activity-based reduction, active data for a storage group with data reduction enabled will skip the data reduction flow for performance optimization. Data not compressed due to activity-based reduction may be compressed later and moved to a compression pool. Data for a storage group with data reduction disabled will bypass the data reduction flow and will be written to the system unreduced.

I/O types include:

• Read: A request to access data that is already populating the array
• Write: Incoming I/O that will consume disk space
• Write-update: Incoming I/O that can change data that is allocated to disk space on the array

The following figure describes the path that the I/O will follow. The path is determined by characteristics of the dataset or the related storage group.

Figure 1.     Data reduction I/O flow for PowerMax enterprise storage systems