Low Bandwidth Replication

Batch-mode Replication

DRBD continuously, and in the PostPath recommended configuration, asynchronously, updates the standby server with changes made to the primary server’s data store. DRBD moves only changes at the disk-block level (if a file is edited, only changed portions of the file have to be propagated). This is highly efficient in most environments. However, there may be some situations – for instance remote offices in developing countries – where bandwidth between remote office and NOC is limited, where a batch-mode approach is more appropriate than a continuous connection.

In batch-mode replication, the NOC server is updated periodically rather than continually. A snapshot technology is used to capture a fixed-moment-in-time-image of the data store at the remote office, and then a file-transfer technology is used to transfer that snapshot to the NOC.

PostPath’s Remote Office Edition includes documentation explaining how to setup batch-mode replication using the logical-volume-manager (LVM) snapshot feature of Linux together with the widely used Linux utility rsync for data transfer. The system is configured so that rsync need transfer only changes since the previous remote office snapshot when replicating. Typically, snapshots can be taken hourly, several times a day, daily, or weekly.

Batch-mode replication can be used in conjunction with Linux-HA and automated failover. In a batch-mode replication environment, if failover should occur users will see their email revert to the most recent snapshot that has been transferred to the NOC. When the remote office server is restored, the system resynchronizes.

Even if you decide not to use Linux-HA with your remote office installation, it may still make sense to run batch-mode replication, so that the data-store is regularly transferred to the NOC where the data can be backed up and managed.