How Distributed Segment Processing Boosts Backup Efficiency in Data Domain Systems

What is Distributed Segment Processing (DSP)?

DSP is a feature of the DD Boost option for Data Domain systems that allows the backup host to perform part of the deduplication work that would otherwise be done entirely by the Data Domain appliance.

Standard deduplication workflow in Data Domain

Segment the data to be backed up.

Generate a fingerprint (hash) for each segment.

Filter out segments whose fingerprints already exist in the system.

Compress the unique segments.

Record references to already‑stored segments and write new data to disk.

How DSP changes the workflow

When DSP is enabled, the backup host takes over steps 1, 2 and 4 (segmentation, fingerprinting and compression). The Data Domain appliance only performs steps 3 and 5 (fingerprint filtering and reference tracking).

The host splits the backup data into 4‑12 KB segments, fingerprints each segment, and sends the fingerprints to the appliance. The appliance compares them with its existing index; matching fingerprints are discarded, while new fingerprints trigger the host to compress the segment and transmit it for storage.

Benefits and considerations

Compression occurs on the backup client, so less data traverses the network, saving bandwidth.

Offloading part of the deduplication workload to the host raises the host’s CPU utilization but lowers the appliance’s CPU usage.

Because the host’s CPU usage increases, DSP should not be enabled on hosts that are already CPU‑bound.