Yeah, this problem has a pretty obvious solution that is implemented by many teams: find a unit of work that we can measure success and failure over, report an aggregate metric over a specified time interval for availability tracking and plotting:

# units successful / total # of units