Storage pools are a subset of physical storage devices in a protection domain. Each storage device belongs to one (and only one) storage pool. The best practice is to have the same type of storage devices (HDD versus SSD or SSD versus NVMe) within a storage pool to ensure that the volumes are distributed over the same type of storage within the protection domain. PowerFlex supports two types of storage pools. For VMS and CV workloads we chose Medium Granularity for use with large file sizes and large read and write block sizes characteristic of these workloads.
- Medium granularity: Volumes are divided into 1MB allocation units, which are distributed and replicated across all disks contributing to a pool. MG storage pools support either thick or thin-provisioned volumes, and no attempt is made to reduce the size of user-data written to disk (except with all-zero data). MG storage pools have higher storage access performance than fine granularity storage pools but use more disk space.
- Fine granularity: A space efficient layout, with an allocation unit of just 4 KB and a physical data placement scheme based on log structure array (LSA) architecture. Fine granularity layout requires both flash media (SSD or NVMe) as well as NVDIMM to create a fine granularity storage pool. Fine granularity layout is thin-provisioned and zero-padded by nature, and enables PowerFlex to support inline compression, more efficient snapshots, and persistent checksums. FG storage pools use less disk space than MG storage pools but have slightly lower storage access performance.
A system can support both fine granularity (FG) and medium granularity (MG) pools on the same storage data server nodes. Volumes can be nondisruptively migrated between the two layouts. Within a fine granularity pool, you can enable or disable compression on a per-volume basis: