I am somehow stuck with my thoughts about an SLA definition, which I plan to determine for a REST-API running on a public cloud.
Imagine having a REST-API service running on Azure which has a database as a dependency. To calculate the SLA for the infrastructure part, I would use the following calculation, regarding that the components are serial.
App Service (Component Availability: 99.95%) --> SQL Azure (Component Availability: 99.95%) = Compound Availability: 99.9%.
But as mentioned at the beginning, I do provide as a service the REST-API for my clients. So looking only at the infrastructure part feels somehow wrong, and I would like to define the availability for the API.
Am I on the wrong path with thinking of it as two parts (application and infrastructure)? And if not, how would I estimate or define the value for a REST-API which does not exist today?