posted

0 Comments

By Ryan Clair, Senior Systems Engineer

The community around Kubernetes is blossoming. Attendance for this month’s KubeCon in Austin doubled last year’s audience to hit more than 4,200 attendees. That impressive number is made all the sweeter due to VMware’s participation in KubeCon 2017 as a platinum sponsor.

So much went down at KubeCon 2017, it would be hard to recap it all. Rather, let’s go big picture and highlight the major themes of the event. Here are three key takeaways from KubeCon 2017:

Takeaway #1: Community

It was extremely refreshing to see such a strong, inclusive community at this year’s KubeCon devoted to helping each other and collaborating together to make the event as great as possible. The community was so impressive that other attendees, such as Matthew (Brender) Broberg, wrote extensively about the topic.

There also seemed to be a good consensus on what Kubernetes is, and what it isn’t. This is important to single out, especially when you make the inevitable comparisons to the OpenStack project. There was a clear sense of what the core of Kubernetes is at the conference, and how a future goal for the tool is making a pluggable framework that satisfies all users.

KubeCon 2017

Another fantastic part of this community is the Special Interest Group (SIG). SIGs facilitate an environment and community where all members can have their voices heard around a specific aspect of an open source project like Kubernetes. It also acts as a starting point for you to contribute to a given project. Sitting in on a Scalability SIG session provided a great opportunity to learn the ins and outs of scaling tests, and unveiled some dashboard gems like perf-dash.k8s.io, which shows scaling tests for different API workloads from 100, 1,000 and even 5,000 node environments.

SIG also showed attendees how small changes in the code in one part of the system can have significant, detrimental impact on other parts of the system. Throughout all of these SIG sessions at KubeCon 2017, there was one common, unifying goal: to make Kubernetes the best it can possibly be.

Best of all, some superstars in the world of Kubernetes attended the event and networked with attendees without any ego involved. Whether it was author Kelsey Hightower or Brad Geesaman, the accessibility to these Kubernetes celebrities really conveyed a strong sense of community around the open source project.

Takeaway #2: Making It Boring

The fantastic Kelsey Hightower kicked off the keynote on Thursday with this paraphrased sentiment: “Well, I was going to go over a Kubernetes 1.9 update, but, as I was reviewing the change log, I realized it’s boring.” That’s arguably the goal for anything in IT—make it all boring so you can add cool stuff on top.

From there, Kelsey gave an insightful live demo on deploying a new Kubernetes cluster (via “OK Google”) and pushing code to GitHub, which eventually made its way onto the cluster without any human input.

The notion of building a software pipeline was definitely in full swing at KubeCon 2017 as well, with many sessions dedicated to CI/CD pipelines. One of the more noteworthy sessions was around  GitOps from William Denniss and Alexis Richardson. The theme was leveraging Git as the single source of truth for everything (i.e. application code and operations , from  infrastructure to monitoring and dashboards) and having an operator (piece of software) that looks at expected state (what was in Git repo) and current state (what’s live and running currently) to ensure the two are the same. This delivers all of the benefits of Git (version control, audit-able, etc.) to all of DevOps.

Takeaway #3: Innovation

The final takeaway is the remarkable innovation flowing through the Kubernetes community. If CI/CD pipeline monopolized a good chunk of the sessions, it’s fair to say Service Mesh was a close second. For a great introduction into Service Mesh, check out this post. Other fresh projects like Isito, Linkerd and Envoy were touched on as well.

In the “Ask Me Anything on Microservices and Service Mesh” session, the prediction was that data plane would become standardized and the control plane would be where the real differentiation will take place. Watching the evolution of a network for services instead of bits will be an exciting arena to watch in the coming future.

Another thought-provoking project was Metaparticle by Brendan Burns. Brendan took “infrastructure should be invisible” to a whole new level, with the idea that instead of treating “Code > Docker > k8s” as three different steps, users should tell the code that it should be sent to Kubernetes as a container. This could be accomplished by importing in the metaparticle library and having it do the heavy lifting for you, using syntax that the developer would be familiar with in their native computer language.

Finally, the Pachyderm session from {code} catalyst Daniel Whitenack highlighted machine learning as a revolutionary technology for data and workflows with an insightful live demo. The complexities of orchestrating all of the various stages in training a machine (some of which require GPUs) and the challenges of versioning the data at scale should not be understated. It’s no wonder then that Daniel created Pachyderm as a tool to address this and extend Kubernetes into new territory.

Interested in learning more about how to operationalize Kubernetes with VMware PKS? Sign up for free webcasts from VMware today! Register here.

Check back into the Cloud-Native Apps blog for more posts on all things Kubernetes, and follow us on Twitter (@cloudnativeapps).