Difference between High Availability data synchronization across local replicas and DAG

Question

I see a significant difference when synchronizing data between Primary - Secondary and between Primary HA and Secondary HA as part of Distributed Availability Groups.

The difference in the amount of data sent across the wire. When making a significant amount of changes on Primary including change tracking enabled tables, we see a huge jump in the amount of data in Redo Queue on all local replicas (synchronous - compression disabled and asynchronous-compression enabled). But across DAG, redo queue is much lower and processing much faster without any significant backlog.

I want to understand on a deeper technical level the difference between local synchronization and DAG.

The specific questions: - What is the difference between synchronizing from Primary to Secondary and across DAG (amount of data) - Can we enable local synchronization to work similar to the one across DAG - What factors that affect the amount of data sent across the wire beside the obvious ones - activity on the DB and HA compression enabled/disabled

Answer 1

What is the difference between synchronizing from Primary to Secondary and across DAG (amount of data)

The main difference is that in a Distributed AGs, there is a "forwarder" which is the primary replica on the secondary side. The data is transferred once across the wire to the forwarder and then sends it to the replicas in its' cluster. My guess is that you've also set the Distributed AG up as asynchronous, which would also show differences over synchronous.

Can we enable local synchronization to work similar to the one across DAG

I'm not sure what you mean, here.

What factors that affect the amount of data sent across the wire beside the obvious ones

Generally it can be traced back to insufficient disks, high latency networks, overloaded VMs, etc. The way that applications are written, to an extent such as the size and scope of transactions can play a large part too. There are too many variables so I listed the main ones.