So Google Kubernetes is a Docker container cluster management solution that helps deploy, scale, schedule and maintenance sets of containers simple.
Apache Mesos is a work scheduler that figures out where jobs (e.g. deploying a Docker container, some batch processing job, etc.) should run.
Kubernetes can be "raw" and run stadalone, or there is the Kubernetes on Mesos option where Kubernetes is deployed as a Mesos framework and runs on top of your Mesos cluster.
I'm wondering what benefits, if any, there are to running Kubernetes on Mesos, as opposed to "raw" Kubernetes?
Mesos can do non-Docker work: it can run Java/Python/Ruby/etc. apps on normal Linux VMs. It can run scripts. It's just a general work engine. So at first glance, it seems that the only advantage Kubernetes on Mesos offers is that it places your Kubernetes cluster onto Mesos, where you can run other jobs that aren't Dockerized.
I'm sure I'm not seeing the "forest through the trees" here, but if I already know I'm going to use Docker and Kubernetes, why might I consider running Kubernetes on Mesos?
I recently gave a presentation at the London Mesos User Group, addressing exactly this question: Can I have Mesos and Kubernetes? with a demo available here. Bottom-line: it is all about hybrid workloads.
If you can and want to go all in concerning Docker, that is, you containerize all your workloads, then nothing (or little) speaks against Kubernetes standalone.