Planning for Deduplication

To implement deduplication in your environment, consider the following factors.

Improving Performance

Deduplication's signature generation and comparison operations require a minimum level of available CPU cycles and disk Input/Output Operations Per Second (IOPS) for successful operation.

Signature generation is, by default, distributed to the individual client computers, but it can be relocated to a more capable MediaAgent if necessary.

Signature comparison is primarily done on the MediaAgent hosting the Deduplication Database (DDB). Locally cached signature comparisons can be used to improve performance in high latency networks.

For more information on performance requirements and locating signature generation and comparison DDB processes, see Deduplication Building Block Guide.

Volume of Data to be Deduplicated

A storage policy copy for a DDB can be tuned for maximum deduplication efficiency by completing the following steps:

  • Configure multiple storage policies, each with a separate DDB.

  • Extend a DDB in capacity through the use of multiple partitions.

  • Seal a DDB and create a new DDB.

For more information on sizing a DDB, see Deduplication Building Block Guide.

Type of Data to be Deduplicated

Similar data type deduplicates more efficiently for storage than dissimilar data types. Dissimilar data types are best assigned to different storage policies for optimum storage efficiency. Similar data types with different data management requirements can benefit from a Global Deduplication Policy.

For more information on Global Deduplication Storage Policy, see Deduplication Building Block Guide.

IOPS Requirement

The minimum IOPS required for DDB disk for extra large, large and medium MediaAgent is:


Extra Large



Suggested IOPS for DDB Disk

20K dedicated Random IOPS

15K dedicated Random IOPS

5K dedicated Random IOPS

For more details, see Hardware Specifications for Deduplication Mode.