My experience at Kubecon Paris 2024

I'll start by thanking Getup, which has kept Kubicast running for 6 years and continues to sponsor this initiative to foster and share cloud-native knowledge in Portuguese with the Brazilian community.

This year we had the pleasant surprise of being accepted into Kubecon as part of the media, an achievement that would not have been possible without the support of Juraci, who is an amazing guy and has been contributing incessantly to the Brazilian community, especially with observability. If you don't know him, run over there for more doses of telemetry. Still regarding this run-of-the-mill experience of being part of the Kubecon media, I had access to various information shared in advance, and I was able to prepare to follow and share everything in the best way possible with all of you. As a bonus to all this, I also had a room reserved for Kubicast for a full day to record several episodes, which served as a meeting point for us Brazilians.

This was truly a week of achievements; I submitted 2 talks to Kubecon, but they weren't accepted, which is common for first-time speakers, in addition to the fierce competition in Kubernetes security topics. However, to my joy, and subsequent despair and anxiety, I was accepted to speak at Cloud Native Rejekts—a fantastic event that happens 2 days before Kubecon, extremely technical, and keeping the community vibes alive. This was the first time I was able to present in English, and I received great feedback. I confess it was not easy at all to face an audience of people who are my community idols and who spoke before me on topics that were very difficult to even follow.

In my talk, I had the opportunity to share about using CEL (common expression Language) in Validation Admission Policy in Kubernetes and how to protect hundreds of nodes at once. This was a chance to also showcase CEL PLAYGROUND, which had already been mentioned the previous day by Marcus Nouble in his presentation on policies and mutating in Kubernetes. If you still don't know about it and would like to learn more, here is a more complete webinar I recently did. (LINK)

To wrap up this long introduction and so many accomplishments, I received recognition as one of the top 50 leaders by Thinkers360. Who would have thought so many things could happen in just one week? This was a pleasant surprise because leadership is one of the skills I consider myself weakest at, but regardless, I am grateful to the expert team at Getup who trust my leadership, whether they are the operations veterans or the GenZ folks at Undistro. A quick note: if you are part of Getup and are reading this text, you are a part of these achievements, I am just the speaker!

Well, the cloud-native world doesn't live on excitement alone, so let's get to the nerd talk!

I want to share some special highlights that haven't been mentioned yet on Kubicast and in the summaries with the group during the conference.

On the first day, or co-located events day, I attended some talks at Multi-TenancyCon, which covered some platform terms and multi-application environments where you have not only users and applications but completely different workloads sharing the same cluster, which requires various configurations, especially those aimed at standardization, security, and isolation. Among these topics, I highlight the focus on security regarding "lateral movement" within the cluster, which can be clearly controlled with Network Policies, already featured in most CNIs like Cilium, which gained major prominence at Kubecon due to the use of eBPF and evangelism by Isovalent.

I also want to highlight here the panel chaired by Kelsey Hightower with Venture Capital representatives for a very good discussion on what they look for in startups to invest in and what they find. It's really worth checking out, I won't dare to summarize, but the statement that stuck in my mind was about sales in a broad sense, whether of your product directly, your career, your achievements, or your team. Here is the presentation link: (LINK)

Of course, I couldn't leave Artificial Intelligence out; this was "THE" big topic of this edition. One highly relevant thing I found was the permanent need for infrastructure, quite different from what many were saying about the possible control and replacement of people by AI; AI still needs a lot of human interaction and support to grow and evolve before reaching T-800. A project that caught my attention was Karmada, which seems to be the 3rd attempt at cluster federation in Kubernetes, and along with getting to know the project, I had the chance to see some cases like Adobe's, controlling several clusters centrally. Nvidia talked about GPU everywhere, and this is still the big question on the topic, perhaps due to resource scarcity or complexity, and that's when the Volcano project stands out. This project manages LLM tasks in the cluster and allows integration with Karmada. An important point that Oracle brought up in its keynote was the use of CPU for AI workloads: going against the flow, it has great arguments like low monetary cost, resource abundance, and, with the popularization of ARM, being able to run workloads with more stable loads and still get a more efficient and sustainable resource use. Among the many AI projects, I'll highlight KAITO (Kubernetes AI Toolkit Operator) here, access it here and learn more. (aka.ms/KAITO)

Another topic that drew attention is the big numbers and how we must prepare for the demand of the cloud-native market. From an approximate value of $547 billion in 2022, the Cloud Native market is expected to grow to nearly 4 times that by 2029, to around $2.3 trillion. A very impressive number that brings some concerns, such as the amount of resources, efficient use of machines, energy, and also the training of the number of professionals required to reach this goal. Regarding the latter, I had a very important chat with Linuxtips here on this Kubicast, live from Paris!

We also had CERN recognized as the largest contributor to CNCF projects, and they shared how they abandoned massive "in-house" software development in favor of contributing to and using open-source projects, realizing that not even they had workloads so unique as to justify that approach. This also brings me to a call for you, the reader, to contribute to the community in whatever way you can. Check out the new "Zero to Merge" program for contributions here: (LINK)

A final highlight is the amount of integrated or dedicated cost optimization solutions. Regardless of the market or workload in question, reducing costs is always a premise, but also a complex issue. Some already well-known projects continue to gain ground, like Kubecost, but we see other platforms like cast.ai and DB systel, which also shared basic tips with the use of VPA to help define resources.

And as an easter egg, I recorded a special closing for you who are reading this article up to this point:

Of course, I wouldn't have been able to cover the whole conference, but I managed to gather an incredible amount of Brazilians there, and we shared MANY insights on Kubicast, so run over there, watch, listen, and share as well!

https://getup.io/en/blog/maratona-kubicast-kubecon-paris-2024

To wrap up, since Kubecon was in Paris, of course, I took the opportunity to stop by the Eiffel Tower—the view from up there is great—visited the Louvre Museum to see some conspiracy theories in paintings and Mona Lisa's passport-style photo, and ran all the way to the Arc de Triomphe to end this endeavor with that feeling of mission accomplished!