When a node joins the subnet it currently syncs from scratch by replaying the whole state of the network from genesis. However, we have checkpoints anchored to the parent chain and a subnet content resolution protocol that we can leverage to enable fast-sync for new joining nodes.
The two mechanisms we are considering are:
The ability to quickly sync leveraging checkpoints and a snapshot of the state in the subnet at that epoch.
Replaying the messages from scratch to allow querying old state
A lazy sync where we sync the latest state and then we "fill wholes" as the state is needed.
The protocol still needs to be spec'ed out and is not implemented in the MVP.
When a node joins the subnet it currently syncs from scratch by replaying the whole state of the network from genesis. However, we have checkpoints anchored to the parent chain and a subnet content resolution protocol that we can leverage to enable fast-sync for new joining nodes.
The two mechanisms we are considering are:
The protocol still needs to be spec'ed out and is not implemented in the MVP.
//cc @vukolic @matejpavlovic for additional input