Wait a minute, how did you handle redundancy in your old solution? Did you do something like 3x redundancy in case some of the nodes went down?