Neat idea. I briefly toyed with the idea of developing blogs or other static content with minimal processing on Lambda. To be clear ahead of time, I don’t know anything about Lambda past what descriptions I’ve seen in blog posts and such. Do tell me if any of this is impossible on Lambda or a Lambda-like service (eg future competitor or homebrew).
In the thought experiment, I’d use a combination of a VM/bare-metal machine (“the Machine”) with Lambda service (“the Service”). The Service would soak up most of the traffic running any computation that could be done right at that point. Aside from forwarding results in simpler form, it might also sync up with the Machine if that sync function itself could happen quickly within its execution window. The monitoring of Service or data for integrity/security would happen on the Machine. If Service runs something like CRDT’s, the Machine might also issue fixes for data to active nodes in Service. That might not be necessary if Machine has a copy of data. If the Machine can upload to the Service, then I could also have the Machine create new, static snapshots of a live blog or some other up-to-date data that it embeds in images for the Service that it uploads. In that way, the Service can also be a data cache for the Machine.
It seemed like you were trying to do everything on Lambda. Did you investigate split architectures like that for databases?
Check out Serverless, a framework for strapping together elastic Amazon services