It is correct that one service application down in a services' cluster would not cause the service itself be offline, but the capability of processing requests will be impact. That is what I said which will impact on the SLA. And also, the reboot may not be predicted, which might be the fatal problem.