Hi Tom

Other reader already told me that I fell into a edge case of Python, with recursion issue. I’m a simple Python practitioner and I don’t know deeply that language. However, recursion is a common programming pattern, and avoiding it in Python might lead to addition complexity or not obvious code architecture. No luck for that test, but it’s still a bad news for python to code without recursion capability.

About Local and Cloud comparison, you can find insight in another article that I wrote about multi CPU on Cloud Run. You have to keep in mind that:

• The number of core are limited on Cloud (1 to 4 with Cloud Run, instead of 8 + HyperThreading for my personal computer)
• The CPU speed is also very different as cloud (about 2.4Ghz, instead of almost 4Ghz for my personal computer)

So, the difference between local and cloud performance are often very different.

About Bare metal, I have any input. However if we talk about serverless, and automatic scalability in general, the boot time (also named cold start) is the most important parameter: it’s the time between a new instance is requested, and when the instance is ready to serve the traffic.
Choose a language that start fast, or a packaging. Java + GraalVM start very fast. Java + SpringBoot start very slow.

If you are on Kubernetes, the boot time + the container size are important. Indeed the lower the container size, the faster the pulling!

For the other criteria, I personally like Go. It’s not the fastest (Java is quicker in my tests), but it’s the most “not bad” language: single binary (no external dependency, single file after compilation), start fast, performance are good, memory management is nice, container size is small, memory footprint is low, developer experience and learning curve are very good, security is built-in, http/grpc native support,…