The first half of this article rings true and resonates with similar co-evolution explanations I’ve been developing, but the example in the second half seems narrowly focused on an optimal use case. It does show that FaaS is a good model for low traffic functions, but doesn’t address the weaknesses around inefficiency at scale or unsuitability for a wide variety of services like stateful backends, caching layers, and multi-object-spanning optimized RPC APIs. FaaS seems like a useful tool to add to the toolbox, but I don’t see it as powerful/flexible enough a layer to completely hide lower layers the same way container platforms do.

For another perspective on co-evolution see these two slides on architecture evolution and FaaS as one of a set of enabling “future tech” pieces at the top of the hype curve.

Slides from my recent POSIX for the Datacenter talk: https://www.slideshare.net/KarlIsenberg/posix-for-the-datacenter