99% of the requests used less than 6.6 ms CPU time, and this stayed flat throughout the evening and the next day. Only in the 99.9th percentile, we saw a little bump which coincided with the donation peak, at around 11.34pm. With AWS Lambda we were used to seeing functions running for 100s of ms, and seconds during the warm-up phase, so having functions consistently run for less than 10ms, without the inevitable penalty, was quite the sight!
A super interesting read and exactly the kind of use-case where I always imagined serverless would shine.
I’m especially impressed by the fact that the team behind We Are Serverless managed to build this in only 4 weeks by focusing on making it as lean as possible.
Not yet personally, I acted as the technical coach/product owner in a “proof of concept” project of the RWTH Aachen where we used a serverless architecture to also make it as lean as possible (see this article for details).
While I haven’t used it personally I really would like to do more with it. I’m convinced that the computing model and the architectures enabled by it won’t disappear but - quite the opposite - will be seen more and more in the future.