How long (ballpark) will a forest recovery take given an environment with 'x' domains and 'y' DC's?
Most of the recovery tasks performed by the Forest Edition take place in parallel so adding more DCs to the process should not significantly prolong the overall recovery. There are a few dependencies however - specifically that the first DC in each domain (the one where SYSVOL is restored primarily) must be completed first before the remaining DCs in the domain are completely restored. If you add DCs to a domain that only has one DC, then it will extend the recovery time by a little as it would need to wait until the first DC was restored. If you add more DCs to that same domain then all additional DCs would be restored simultaneously after the first DC completes and would not add significant time to the recovery.
The two actions that take up 90% of the recovery time are:
For #1 - there are 3 reboots required. The 1st reboot is into DSRM mode, the 2nd reboot is back into normal mode after restore from backup has been completed and 3rd reboot is after the Forest Edition resets all passwords (computer, trust, Kerberos), increases RID pool and enables replication.
For #2 - the restore from backup will be quick in a test environment but could be quite long in a production environment with a large AD database. The best way to determine how long it will take to restore the AD database is to perform a restore of the customer's database in a lab environment.
The time to restore a DC can be calculated using the following formula:
DC.recovery.time = (3 * dc.reboot.time + offline.restore.time) * 1.1 (for pre/post-recovery tasks)
The other factors that will affect the overall recovery time are:
The important thing is that adding additional DCs to the mix does not multiply out the recovery process unless they add incremental time because of the factors listed above (particularly the last two).
