Gregory's ServerWhy is "Cost per Request" not a common metric in every #observability strategy?
 2 comments
@garritfra observability is just about the ability to ask questions. To me, knowing the cost of the application is a great question to ask, especially as we get to be able to see that data better with RUs for database services and charges for lambda and functions. We monitor that extensive with our lambda implementations. |
@garritfra Cost is at best a synthetic metric, and can only be calculated with a full context.
Take for example a request that asks for the top K results in the past month. The cost is different if the database contains 1000 or 1M results. Similarly, the cost of a request that returned data after 10s may be lower than the cost of a request that times out after 5s, if UX is impacted.
Typically, instead of calculating cost one would calculate business continuity impact. This is what SLI does.