As I sit and write down my thoughts on KubeCon, the first thing I can think of is “Wow!” This show was amazing on so many levels as there were a number of great talks and keynotes, new companies showing up in the ecosystem, and a general feeling of excitement as cloud-native growth continues to accelerate. Here is my color on the conference, ecosystem, and some of the more interesting conversations I had.
Find me on Twitter (@nirajtolia) and let me know what you think!
Conference Growth: I have been attending KubeCon since 2016 and the growth this year has been just amazing. We had 12,000 attendees show up and this is a large jump from only 1,000 attendees in 2016. Based on our conversations from the show floor, a large number of folks are in production with Kubernetes at scale but a significant part of the audience was also new to both Kubernetes and KubeCon + CloudNativeCon.
We loved this mix of attendees here as they were split between end application developers and users, infrastructure operators, administrators, and vendors. Unlike OpenStack conferences back in the day, the number of vendors seemed to be a relatively small part of the conference. The “fresh blood,” in particular, was very encouraging to see as it is a true indicator of usage growth. These new members of the cloud-native community were predominantly here not to kick the tires but, more importantly, because they saw Kubernetes going in production in the next few quarters at their companies and they needed to be prepared.
Finally, while there is always the risk of marketing and hype outpacing reality, I think things are fairly balanced with an optimistic outlook given the above trends and a self-aware and open communitity that is constantly iterating in response to constructive feedback.
Talks and Keynotes: The conference had a large number of “Day 0” events (including Cloud Native Storage Day that we organized) and a packed agenda of keynotes, talks, project updates, and more! Check out the agenda if you haven’t already done so as there is something that caters to every audience. The talk slides and video recordings are also available now.
Attendees did seem to have constructive feedback about some of the CNCF project update keynotes. The updates felt forced and, at times, felt like a plug for the commercial company backing the project. However, this can be easily fixed by making these presentations focus more on use case studies for the projects, live demos, and the end-user impact. Finally, given the large percentage of new members joining us at every conference, I think a focus on adding more intro-level talks to the agenda (or even a dedicated track) will serve the community well. After all, even with all this growth, it is still early days for the CNCF!
There were also conversations around the lack of major product announcements but I believe that is because folks are only used to vendor conferences being this large. The community and project-driven talks set the right tone for this conference.
The CNCF Ecosystem
Mirantis/Docker: The news of Docker Enterprise (including the Docker Engine and CLI) being sold to Mirantis was announced before KubeCon but it was definitely a conversation topic at the show. We aren’t here to delve into Docker’s missed opportunities or talk about how impactful the Docker project has been though. Instead, it is interesting to analyze what Mirantis is trying to do.
Mirantis, as it moves away from OpenStack, has been trying to enter the Kubernetes market since early this year but didn’t make traction compared to the larger on-prem Kubernetes distributions (Red Hat/IBM, Pivotal/VMware, and Rancher). This purchase gives Mirantis some instant brand recognition and credibility in the market but how well they leverage this will be seen in 2020.
The Container Storage Interface: I am admittedly biased here but we saw the Container Storage Interface (CSI) being talked about everywhere with folks deploying more stateful applications and storage vendors. There were also Day 0 events, the SIG Storage F2F meeting, and talks that showed how users could leverage CSI (check out this talk by VMware as an example for CSI-enabled data protection).
The increasing popularity and acceptance of CSI as the one-true-way to provision storage in Kubernetes has been amazing for storage administrators and users. They now have greater flexibility to pick the best (or mix multiple) vendors for the environment they are in and increase application portability. However, the increased use of this abstraction is also making it hard for storage vendors that have pivoted (selling traditional storage is hard!) into the cloud-native space to compete. Definitely a market dynamic for us to keep an eye on.
Vendor Ecosystem: Just like the number of attendees, the number of sponsors at the conference was also at an all-time high. As more companies move Kuberetes into production, there are a large number of companies working on solving hard “Day 2” customer challenges in this new cloud-native infrastructure space. We do predict some upcoming consolidation as there are a large number of cloud, enterprise, and startup entrants in the “Day 1” space (e.g., with multi-cloud Kubernetes cluster management or Kubernetes distributions). At the same time, there are still open needs in the area of simplicity and scalability tools for Day 2 operations (e.g., security, data management, developer tools) and you should expect to see continued activity in this space.
Controversy Around Technology
Linkerd vs. Istio: Like recent KubeCons, there was some angst around service meshes and competition within CNCF. On one hand, you have Linkerd, a CNCF project and a service mesh that has seen significant production use. On the other hand, Istio, a non-CNCF project that Google and IBM are heavily investing in, seems to get more airtime but, for now, has the reputation of being complex and heavy weight. Istio, thanks to its increased visibility, has also become synonymous with service meshes as far as our discussions with customers are concerned. How CNCF navigates this in the future will be quite interesting.
Is YAML Bad For You?: As exemplified in my below tweet, there was some (IMO misguided) backlash against the use of YAML at KubeCon. I discussed this with a few people as I think YAML is actually great and, while Kubernetes might overdo it a bit, it gives folks a lot of power and is actually the right thing to do and stick with.
Unpopular opinion: I think YAML is actually great and, while #Kubernetes might overdo it a bit, it gives folks a lot of power and is actually the right thing to do and stick with.— Niraj Tolia (@nirajtolia) November 22, 2019
As Bryan Liles pointed out, YAML can be quite dense and, when faced with reams of infrastructure-as-YAML-code, it can become opaque and hard to decipher. However, I would argue that the average Kubernetes user probably sees less YAML on a daily basis compared to folks who are much closer to the Kubernetes core. The information density actually helps there (it is a feature, not a bug) and I believe that the success of the ecosystem is partially derived from the ability to concisely use YAML to express the right abstractions. This nuance is being lost and is being translated to “YAML is bad.” While you can always have too much of a good thing, let’s not throw out the baby with the bathwater here.
See You In Amsterdam!
Overall, as you might have observed on Twitter, we were thrilled to be a part of KubeCon and talk to everyone that showed up to say hi and learn more about what we are doing to make Kubernetes backup and mobility easy, secure, and extremely simple to use. We are equally excited about seeing more of you in Amsterdam. Till then!