measuring service reliability

I am part of Servers service support team in a large company. we have been measuring our performance by the percentage time that we have the services(servers) available. We have been having problems defending our availability figures. I have the opinion that we could rather use service reliability to measure our performance. But how can that be measured????