I mostly agree – which probably betrays my age – though when the load exceeds one server, you're going to have to orchestrate no matter what: deployment, load balancing, fail-over, system updates. While I'm critical of Kubernetes' complexity, I still think that for some people, having that orchestration handled by a tested layer instead of implementing it from scratch in-house is still going to be a benefit.

After working with and securing large bare-metal and monolithic virtual host environments for way too long, there's got to be a better way to deploy services and update systems than the snowflake hell some monolithic approaches seem to bring.

Perhaps there's a middle way between both worlds that has yet to gain any momentum.