A key attribute of what constitutes a *serverless service* is that it should offer a pay-per-use (aka PAYG) pricing model. Specifically, the metrics used in determining the price for a [[Serverless MOC|Serverless]] service should all be based on how much the service consumer is using the service. ## Pay-per-use metrics Examples of such metrics include: - Number of requests or invocations - Amount of data stored ## Non Pay-per-use metrics The following metrics DO NOT follow a pay-per-use model: - Instance size - Hours an instance is "available" to serve requests Also, if a service is not being used at all, it should **scale to zero** and users should not be **paying for idle**: > A Serverless solution is one that **costs you nothing to run if nobody is using it** (excluding data storage). [^pj1] [^pj1]: [A simple definition of “Serverless”](https://pauldjohnston.medium.com/a-simple-definition-of-serverless-8492adfb175a) by [[@Paul Johnston]] --- ## References - https://serverlessfirst.com/serverless-definitions/ --- tags: #Serverless