This choice will depend on the business objectives and complexity of measurement, but in most cases, organizations consider only critical services for such KPI (measuring noncritical services only adds effort without bigger benefits).
Regarding calculation, in general, it is considered the total sum of interruption times in a given period (e.g., day, week, month, etc.), where the downtime periods are monitored either by monitoring systems or by reports sent by users. Both approaches have their advantages and disadvantages, which should be evaluated considering the business context.
This article will provide you a further explanation about monitoring: