This is the second post I see about this edge functions, and while I appreciate the advantages (decreasing network latency should definitely have an impact), will this matter much if the database is still centralized? If most edge functions still need to talk to a centralized database, won’t that be essentially the same, performance wise?
Yes, until we have actual edge storage, the use cases are more limited, but "edge functions" can be used for:
* Stateless / static use cases.
* Optimizing use of edge cache.
* Intelligently redirecting traffic to geographically-diverse back-ends -- including other cloud services that are already globally distributed.
For example on the last point: If you want to use Google Spanner as your database but you need to put business logic on top of it, then where does that business logic go? If it goes in a VM, either you have to make sure to replicate that VM all over the world (can be cumbersome and/or expensive) or you've lost Spanner's advantage of being globally distributed. You could put your logic at the edge, and then it will usually sit between the user and the nearest Spanner node to that user, which is ideal.