This year at VMworld I had the pleasure to co-present with my fantastic colleague Pranshu Jain a session I always wanted to deliver to our VMware communities: Kubernetes Resource Management for vSphere Admins. Initially we did not want to make it a deep dive to also attract newcomers to the world of Kubernetes. But after the first internal reviews and stats on the session signups for VMworld US we decided that it actually is deep dive content.
That was the right choice, because VMworld participants love deep dives and the knowledge on Kubernetes compared to the previous years has grown a lot. So we could assume some basic knowledge on Kubernetes for the session. And since the presentation was using a lot of vSphere constructs participants were able to relate the concepts discussed.
After a short VMworld 2018 recap and a quiz to get everyone excited and interested in what’s to come in the session, we covered what resource management is, why it’s important and why you need to think beyond the layer you own/manage/operate. Turns out it’s a pretty complex topic where a lot of things can (and will) go wrong. And just because you follow a best practice for your platform, that does not automatically translate into a best practice for the upper/lower layers involved in delivering a full software and application stack.
If you are a vSphere and/or Kubernetes platform administrator (architect), you definitely want to understand how the different platform layers interact to build a scalable, robust and efficient application platform. To simplify the understanding of all the moving pieces, we covered the lifecycle of a pod (application) in Kubernetes from a resource perspective. That includes pod definition, admission control, scheduling and enforcement (execution). Since this was a deep dive, we could also cover container internals such as Linux Kernel cgroups used by Kubernetes to mode resource requirements and best practices for aligning them with vSphere constructs (resource requests/limits).
The feedback (scores) we got for both sessions (US/EU) were outstanding. And as discussed during the presentation, we do take them seriously.
So a BIG THX for everyone who attended, especially the one in Barcelona where we got the last slot on the last day - with some incredible presenters like Ben Corrie presenting deep dives on Project Pacific in parallel.
You can watch the sessions for VMworld US/EU via the following links:
If you have any feedback on the session and material/concepts discussed, feel free to reach out to me on Twitter. I’m not sure if I’ll try to submit a similar session again for 2020. But depending on the feedback and if this content is valuable I might reevaluate my decision :)