Agree. With AWS, once you start using multiple services (e.g. lambda, S3, dynamo db, etc), it's difficult to estimate/model the costs in advance. First you ha

Agree. With AWS, once you start using multiple services (e.g. lambda, S3, dynamo db, etc), it's difficult to estimate/model the costs in advance. First you have to estimate how many times your app is pulling down a file or hitting a given db table. But this is dependent on unpredictable user behavior. You don't really know what users will do with your app until they actually do it. Then, with e.g. Dynamo DB, you have to think about how many indexes you have on a table, and how often your app will query those vs how many times your app might do a full table scan. You have to allocate these specially defined AWS resource units. I don't find it intuitive. I wonder how many developers really know in advance what their app is going to cost per user.