You need to multiple that out by request servicing time. Say your application uses 25ms real CPU time per request. That's 40 reqs/sec/cpu core. On a 4 core ser

You need to multiple that out by request servicing time. Say your application uses 25ms real CPU time per request. That's 40 reqs/sec/cpu core. On a 4 core server, that's 160reqs/sec. That's 625 seconds to clear that backlog assuming a linear rate (it's probably sub linear unless you have good load shedding). So that's 10 minutes to service 100k requests in your example. I'm ignoring any persistent storage (DB) since that would exist with our without Lambda so that would need its own design/architecture.