Source side deduplication is a mechanism within Protector that improves network bandwidth utilization by avoiding sending the same data multiple times.
In simple terms, as soon a block on the source machine is detected as changed during post-scan in a batch policy, those blocks are transferred to the repository. Source side deduplication uses Single Instance Store (SIS), a one-by-one method of scanning/transferring data from source to repository during initial synchronization and subsequent resynchronizations. If, for example, twenty 5TB source nodes (each having much of the data duplicated across them) are backed up, the repository will ingest data from the first source node, go to the second source node and compare its data with what is already ingested, then transfer only non-duplicate data. This process is repeated with the remaining source nodes.
Even if some duplicate files get through, with source side dedupe enabled, SIS post processing will remove those duplicates. SIS post processing is applied to the dataset once per day during repository cleanup.
The changed block transfer method is far more efficient and has less impact than traditional backup systems.