Ask HN: How do you monitor your AWS costs?

This is the question for most cost/time services. Most of them refuse to put in a real time meter or hard stop after x usage. It goes against their business model. They much rather argue overage than limit usage from their side.

Early and often. When designing a system, you have a choice to use AWS for all of it, some of it, or none of it. You need to calculate the tradeoffs of AWS (increased price) over other solutions. Just like you think about which database to use, or how to cluster components, you should also think about how much deployment will cost.

"How much will this cost" should not be calculated as an afterthought, but rather from the beginning, at the same time you're answering all other architectural questions.

If you've calculated the cost from the beginning, you may as well track costs in your monitoring code and metrics system. For example if you've had 10 VMs running the last hour, log the price of them to your metrics system. Then you can graph prices along with any other metric and see how bottlenecks drive cost.

Try to anticipate the bottlenecks of your system, and see if there's an easy way to avoid AWS for those components. That is, if you're an image host, you probably don't want to pay per GB of bandwidth at AWS when you can pay per capacity at any dedicated/colo provider.