More than 9,000 people convened at the Salt Palace Convention Center in Salt Lake City for the first day of KubeCon + CloudNativeCon North America. The mood was energetic and lively and the audience was primed to dive into the themes of the day: artificial intelligence and platform engineering.
Here’s a look at some of the highlights.
Keynotes
Welcome, and banish trolls!
The conference began with a video welcome from Priyanka Sharma, executive director of CNCF, who reminded the audience that they are the people who built cloud native.
Chris Aniszczyk, CTO of CNCF, continued with this theme, pointing out that nine-year-old CNCF now has over 200 projects with over 255,000 contributions across 193 countries. That amazing level of success has also drawn attention to the open source community, and not all of it the right sort of attention.
In some cases, so-called patent trolls, whose sole interest is buying patents and threatening adopters of technology with lawsuits, have started coming after the open source community with these frivolous claims. As Sharma and Aniszczyk stressed, the time has to come to stop this.
Jim Zemlin, executive director of the Linux Foundation, took up that thread, vowing to crush patent trolls, promising “no negotiation and no settlements.” (There were definitely cheers of support from the audience at this point.) Zemlin pointed to the Linux Foundation/CNCF partnership with Unified Patents as already showing a 90% success rate in defeating patent trolls.
The relationship with Unified Patents has already invalidated dozens and dozens of patents, explained Joanna Lee, VP of strategic programs and legal at CNCF. But, echoing what Zemlin said, she said a broader push is needed because patent trolls are responsible for more than 80% of high tech litigation today. Lee said that was one of the driving forces behind the decision to launch the Cloud Native Heroes Challenge, where cloud native developers and technologists can earn swag and win prizes by helping protect our ecosystem from patent trolls. (And again, there was definitely audience applause at this announcement.)
Multi-cluster batch jobs dispensing with Kueue at CERN
Marcin Wielgus, staff software engineer at Google, and Ricardo Rocha, lead platform infrastructure at CERN, took a deep dive into how Kueue tackles one of the trickier issues around admission to Kubernetes workloads. Working around the principles of “quota borrowing and fair sharing,” Kueue helped CERN and its data physicists improve particle flow by having a central place to submit jobs.
Take a peek under the hood of cloud native AI at scale
One of the biggest challenges running AI workloads is that they tend to live in black boxes, so it can be difficult to know what’s really going wrong, and that’s what CoreWeave’s Peter Salanki, chief technology officer, and Chen Goldberg, senior vice president of engineering, set out to demonstrate. Through a series of steps and processes, they walked through how CoreWeave was able to cut “interruptions” time in half and save money by making sure it was possible to look under the hood at what was actually going on.
Their motto: Failures are inevitable but management is key.
Paving the way for AI through platform engineering
If there’s a secret sauce to AI success, Kasper Borg Nissen, staff engineer at Lunar, would argue that it’s applying platform engineering principles to AII platforms. Classic platform engineering concepts – self service-ing, explicit APIs, paved paths, platform as a product, etc. – can be tweaked to work for an AI journey. That worked at Lunar where right now over 60% of text communication with customers is now handled by AI without human intervention. That translates into a 93% reduction in support time resolution. Or to look at it another way, AI has taken on the equivalent of 13% of full time staff work. And Lunar did it using existing cloud native investments.
Nvidia case study – many facets of building and delivering AI in the cloud native ecosystem
Chris Lamb, vice president, computing software platforms at NVIDIA, walked attendees through NVIDIA’s long history in the cloud native/open source ecosystem. Lamb invited the audience to explore digital human architecture using Blueprint to create a digital human assistant. Try a demo with James.
The engineering future of generative AI platforms on Kubernetes
Aparna Sinha, senior vice president and head of AI product at Capital One, thinks generative AI has the potential to be as mainstream as the internet or mobile technologies. One huge advantage it has? It comes with a more natural human interface immediately giving it a broader range of users thanks to audio, images and video. GenAI has hundreds of use cases today just in banking, Sinha said, from automating coding to processes across the back office. “AI is the next layer,” she said. “Let’s build the puzzle together.”
All of the KubeCon fun
For all of the technology and networking and learning going on at KubeCon, there is also plenty of fun to be had! Attendees can try the local speciality dirty soda (it’s delicious and available in a wide variety of flavors both regular and diet), try their hands at pickleball and curling as well as arcade and console games, pet some cute dogs, visit the relaxation station, build with popular building blocks, and the try the coffee bars…and these are all available within the convention center.
Session snapshots
Better pod availability – a survey of the many ways to manage workload disruptions
Zach Loafman, staff SRE at Google, made the case for a new taxonomy for disruptions – basically breaking it into bad and good disruptions, then walked attendees through a case study in pod disruption. His takeaway: managing pod disruptions on Kubernetes requires a series of tradeoffs and it’s important to think about the cost of disruption.
Can your Kubernetes network handle the heat? Building resilience with AI chaos
Surya Seetharaman, principal software engineer at Red Hat, and Lior Lieberman, site reliability engineer at Google, outlined the challenges involved in Kubernetes networking and suggested that even being proactive is probably not sufficient to get ahead of the issues. Their advice is to embrace chaos testing and use CNCF projects including Litmus, Krkn, Chaos Mesh and kube-burner. But, because human-created chaos isn’t enough, use AI to enhance the chaos experiments.
Announcements from Day O and Day 1
Scalable authentication across organizations with Keycloak 26
Announcing the release of KubeVirt v1.4
Announcing the Cloud Native Heroes Challenge
Announcing the inaugural contest for the Cloud Native Heroes Challenge
LitmusChaos gains adoption in lower environments
Cloud Native Computing Foundation Announces cert-manager Graduation
Cloud Native Computing Foundation Announces Dapr Graduation