> Actually that's not true. We aren't using a system like SimpleDB > where your changes might not be immediately available in subsequent > queries by the same user.
Isn't the point of replication to handle massive reads? If so, you'd be using load balancing? So you cannot predict which server will be chosen?
