Data Domain replication can divided into two types, one is Data Domain Native Replication which is controlled by Data Domain; another is Managed File Replication in which is controlled by the backup software.

If you want to use DD replication, you need to install replication license at the source DD system and target DD system.

DD replication feature is more efficient, because

1)     It only copies the deduplication and compression data.

2)     In the replication, source device will send Index of replication data to the target device. If the index already exists, the target device will create a file and refer to the existing index; if the index does not exist, the source device will send replication data segment to the target device.

3)     Sending the index only takes up a small amount of network bandwidth.

Data Domain Native replication has three replication types, which are Directory, Mtree and Collection. They are copying the entire directory, or the file system. Once configured and initialized, the replication will automatically run.

1)     Directory Replication copies subdirectory under /data/col1/backup;

2)     Mtree Replication copies other Mtree except /data/col1/backup;

3)     Collection Replication copies the entire file system, including /data/col1/backup and other Mtree.

Note: On the DD GUI, VTL copies are called Pool Replication. If VTL uses backwards compatibility mode, the tape path is /data/col1/backup/vtc/ / , so the copy is a Directory Replication. If VTL uses Mtree mode, the tape path is /data/col1/ /.vtc/ , so the copy is an Mtree Replication. All Pool Replication in the DD command line appears as the actual type of replication (Directory Replication or Mtree Replication).

Managed File Replication has a variety of names, such as called SLP (Storage Lifecycle Policy) in NetBackup, called the CCR (Clone Controlled Replication) in the NetWorker. The replication object is one of the file under DDBoost Mtree. After the configuration, replication initiated by the backup software and DD completed the replication and notify the results to backup software.

DD Replication supports many replication topologies, including one-to-one (A->B), bi-directional (A->B, B->A), one-to-many (A->B, A->C), many-to-one (A->C, B->C) and cascaded (A->B->C).

Directory Replication and Mtree Replication support source and target have different DDOS version (For example 5.0.x and 5.2.x), but cannot over cross 2 versions. Collection Replication requires source and target running the same DDOS version. Managed File Replication requires that the target using the same or higher DDOS version than source.

Directory Replication

1)     Through replication logs to records the operations of source replication directory, including create, write, modify and delete.

2)     The system will check the replication source directory every 15 minutes to verify whether there are changes or not.

3)     In phase 1 of 3 of initialize and resync, source directory cannot be written.

4)     In the resynchronization, the target device needs to create snapshots for the existing data of the target directory and delete all existing data. If the existing data are large numbers of small files (millions or more), resync time will be much longer until removal is complete (up to several hours).

5)     The same device can be configured to more than one Directory Replication.

Mtree Replication

1)     Point-in-time source mtree data is recorded by snapshot and this snapshot is sent to destination mtree on destination DD.

2)     When all snapshots are sent, system creates a snapshot for source mtree in every 15 minutes.

3)     If the last snapshot took more than 15 minutes to be sent, then the system creates the next snapshot immediately upon the send completion. If no, then the system creates the next snapshot at the time of 15 minute after last snapshot creation.

4)     If there are user-defined snapshots, then the system sends the user-defined snapshot immediately after the previous snapshot was sent. When sending user-defined snapshot, system stops creating the snapshot.

5)     During phase 1 of 3 in initialize or resync, source mtree is writable.

6)     One DD can have multiple mtree replication contexts configured.

7)     The data amount of pre-comp remaining is inaccurate during initialize or resync, and it will be accurate upon the completion of initialize or resync. For example: The pre-comp remaining could be 10TB, while 1TB data replication is being initialized or resynced.

Collection Replication

1)     The content to be sent is based on block-level change.

2)     If any block of file system is changed, then the source DD sends the change to the destination DD immediately.

3)     Cannot be resynced.

4)     One DD can have only one Collection Replication context configured.

Manage File Replication

1)     Defined and launched by the backup software.

2)     Data Domain determines the execution and return of the results.

Note: Currently, on DDOS versions, if the backup software stops Manage File Replication, the backup software will immediately record the replication as failed. But on the Data Domain, replication will continue to run until all current file transfer is complete. And it cannot be stopped via command line or GUI. If you want to immediately stop Data Domain from copying the file, you can consider restarting the Data Domain file system or turnoff DDBoost function (Turnoff DDBoost feature will affect the entire file system and DDBoost functions, so we do not recommend this function). If replication is complete, the backup file on the target device cannot be used by the software directly.

Data Domain Replication Throttle

1)     Data Domain replication using network bandwidth can be limited by Data Domain Replication Throttle.

2)     The previous version of DDOS 5.5, all replication pair/context on the source device using the same Throttle.

3)     Starting from DDOS 5.5, the replication pair/context of the different target can be configured different Throttle.

4)     Throttle requires configuration at the source device, but if the Throttle of the target device is set to 0 (disabled), then the replication pair/context to this target device will stop.

5)     Throttle can used for Data Domain Native Replication (Directory/Mtree/Collection) and Managed File Replication.

6)     Throttle entered into force physical network interface of Data Domain. If the network interface for replication use virtual network interface, the Throttle should be reduced.

Best practices:

1)     The Backup to Data Domain and Data Domain replication are best used with a different network interface.

2)     For Mtree and Directory Replication, in order to improve efficiency, please spread data to multiple replication pair/context.

3)     If the network bandwidth is less than 6Mbps, you can enable Replication Low-Bandwidth-Optimization to reduce the amount of network data. The changes will take effect in the next time the file system cleaning is completed. But after the changes, the first clean cycle will substantially increase, such as from a few hours to more than 10 hours.

The replications type Fig: