Planning for deduplicated volumes

A deduplicated volume or volume copy can be created in a data reduction pool. When you implement deduplication, you must consider specific requirements in the storage environment.

Deduplication is a type of compression that eliminates duplicate copies of data. Deduplication of user data occurs within a storage pool and only between volumes or volume copies that are marked as deduplicated. However, there is no requirement for all nodes in a system, and therefore all I/O groups, to support deduplication. You can create deduplicated volumes in an I/O group when no compressed volumes or volume copies are in regular storage pools (that is, when Random Access Compression Engine (RACE) compression is in use on that I/O group). DRP compressed volumes can coexist with RACE compressed volumes in the same I/O group.

You can migrate any type of volume from a regular storage pool to a data reduction pool. You can also migrate any existing RACE compressed volume to a data reduction pool. After you migrate a volume to a data reduction pool, you can then create a deduplicated volume.

The following software and hardware requirements are needed for deduplication. There are also update and performance considerations.

Code level 8.1.2 or higher is needed for data reduction pools.
Code level 8.1.3 or higher is needed for deduplication.
Nodes must have at least 32 GB memory to support deduplication.
Nodes that have more than 64 GB memory can use a bigger deduplication fingerprint database, which might lead to better deduplication.
You can use the FlashSystem A9000 deduplimator tool to estimate how much capacity you might save if a standard volume that a host can access was a deduplicated volume.

RACE compression and deduplication are not supported in the same I/O group. However, data reduction compression and deduplication might be supported on certain platforms. Table 1 details the features that are supported on each platform.

Table 1. Supported compression features
Product	Platform	Node/canister memory (GBs)	Supported features
Product	Platform	Node/canister memory (GBs)	RACE	DRP	Compression	Deduplication
SAN Volume Controller	2145-SV1/2147-SV1	64/128/192/256	Yes¹	Yes	Yes¹	Yes
	IBM Spectrum Virtualize™	Any	Yes (dual CPU only)	Yes	Yes²	Yes (greater than 32 GB required)
	2145-DH8	32/64	Yes¹	Yes	Yes¹	Yes

Note: RACE and data reduction compression is assumed to be supported within the same I/O group at the same time, unless the platform does not support compression or the matrix indicates otherwise.

1 - Requires compression hardware for new clusters.
2 - DRP compression is supported by single or dual CPU. Dual CPU RACE restriction still applies. Does not support DRP compression and RACE compression in the same I/O group at the same time.