I'm running monerod on an Amazon EC2 instance with an SSD volume (gp2 volume type), and synchronization from scratch gets extremely slow after I've used up my burst credits (allowing me up to 3000 IOPS for a limited time instead of 100). See also my question here for details.

My follow up question now is if PR #1506 will significantly decrease disk IOPS as a side effect of addressing issue #1463. The description says that the number of DB txn is reduced to 1 for each batch, and I'm assuming but am not entirely sure that this also means much fewer IOPS on the disk side.