CONFERENCE SUMMARY

KubeCon Chicago Key Takeaways

The (slow) rise of AI, the domination of platform engineering, and the refocus on developer experience at KubeCon NA 2023

Daniel Bryant
10 min readNov 12, 2023

This KubeCon NA was a very different experience for me, as I wasn’t representing a company or running a booth. However, I still had a blast, and it was amazing to catch up with so many people from the cloud native community. It was also great to be back in Chicago, and the city looked after us well — I had forgotten how much I enjoyed deep-dish pizza!

Here are my key takeaways from KubeCon NA 2023:

  1. The cloud native community is embracing AI/LLMs… slowly
  2. DevOps is so passe: Platform Engineering all the things!
  3. There’s gold in them thar hills! Selling picks and shovels in the platform rush
  4. Kubernetes should remain unfinished (and evolving)
  5. Don’t forget about developer experience!
  6. Increased focus on app development and integration
  7. Cloud native communications: Bundling FTW?
  8. Security is big business
  9. More focus on sustainability: Observability, scaling & FinOps
  10. Community, community, community
Developer experience was highlighted in the keynotes

I want to say a big thank you to the CNCF for organizing the event and also a huge thanks to all the speakers, sponsors, volunteers, and attendees. You all rock!

Setting the scene: Here comes the late adopters

The general vibe I got from attending KubeCon Chicago was that we’ve very much moved into the late adopter phase of cloud native technologies.

Of course, there’s always the danger of getting stuck in a (cloud native) bubble with conferences like KubeCon; it would be easy to walk away from the event thinking the world runs on Kubernetes. This isn’t true, and “Gartner Says 50% of Critical Enterprise Applications Will Reside Outside of Centralized Public Cloud Locations Through 2027”. However, if you squint, the fact that almost 50% of workloads are running (or will run) in the cloud signals the arrival of the late adopters.

Source: Diffusion of Innovation, Wikipedia

I sensed this shift in the event keynotes and the value propositions pitched in the vendor hall. More emphasis was placed on reliability and trustworthiness over innovation, product suites and ease of integration over “best in class” point solutions, and partnerships and collaboration over trailblazing individual organizations.

With this in mind, let’s dive into the key takeaways more deeply!

The cloud native community is embracing AI/LLMs… slowly

In stark contrast to the OpenAI DevDay held on the same day as the KubeCon NA co-located events, there wasn’t a lot of AI-themed content at KubeCon. If you haven’t watched the OpenAI DevDay keynote, I encourage you to do so to get a glimpse of what’s to come. Ed Sim also provided a great roundup in his weekly What’s Hot in Enterprise IT/VC newsletter, and my InfoQ colleague, Andrew Hoblitzell, shared his thoughts in “OpenAI Announces New Models and APIs at First Developer Day Conference.”

Granted, the first day’s KubeCon keynote led with the topic of AI, but this felt a little “bolted on” and more focused on providing infrastructure for AI/LLMs rather than using this in development or operational workflows. I won’t be too judgemental here, as an event like KubeCon takes months of planning, and the developments in AI are happening at a daily pace.

A couple of vendor booths focused on “ChatGPT for the cloud”, but most of my chats with sponsors were firmly focused on the here-and-now of building platforms. Andrew Fong, CEO of Prodvana, nicely captured my thoughts on this with his “AI essentially does not exist at KubeCon” comment on LinkedIn:

If you’re interested in learning about the potential for AI and LLMs in the platform space, I explored this topic with several cloud native and DevOps leaders, including Helen Beal, Abby Bangser, Matt Campbell, and Steef-Jan Wiggers, in a recent(ish) InfoQ Cloud and DevOps Trends Report podcast.

DevOps is so passe: Platform Engineering all the things!

I’m sure we’ve all seen the Minions/Gru platform engineering and DevOps meme by now (if not, I mentioned it in my KubeCon EU 2023 summary), and everyone was doubling down on this at KubeCon. If you search the program schedule, you will find at least 2x the number of talks that mention platform engineering compared to talks that mention DevOps. Practically no talk led with DevOps in the title, even though they all channelled the spirit of DevOps.

I think this is a natural evolution of the space, which comes with all the good and bad associated with this type of change.

The “good” is that we are forming a community around the topic and crystalizing best practices within the industry. I’ll give a big shout-out to the Cloud Native Computing Foundation (CNCF) TAG App Delivery team for publishing their “Platform Engineering Maturity Model” (and a special shout out to Abby Bangser from Syntasso for herding the cloud native cats)

There will also be a Platform Engineering Day co-located event at KubeCon EU 2024. We’ve come a long way since I collated a bunch of ideas about platform engineering emerging from the community and presented this at KubeCon 2022 as “From Kubernetes to PaaS to … Err, What’s Next?

There’s gold in them thar hills! Selling picks and shovels in the platform rush

The “bad” part of everyone adopting platform engineering is that many vendors are in danger of “platform washing” their products. Granted, many of the existing CNCF landscape-inspired products were helping folks build platforms before this was officially a thing. But I’m not sure everything I saw marketed as platform engineering-friendly at KubeCon actually was. This means that end users must look more closely when buying products to ensure they meet their needs.

If a tool doesn’t integrate well with the existing cloud native stack, hasn’t been built to address clear use cases (i.e. solves a “job to be done” problem), and can’t be used in a self-service manner, it’s probably not been built with the concepts of platform engineering in mind.

Of course, in any (platform engineering) gold rush, there’s always money to be made selling picks and shovels, and technologies like Crossplane and KubeVirt kept popping up on my radar:

Kubernetes should remain unfinished (and evolving)

I thoroughly enjoyed Tim Hockins’ final day keynote, “A Vision for Vision — Kubernetes in Its Second Decade”, in which he opined about the value of making tradeoffs in the near term to benefit the longer-term vision.

It would have been easy for the community to attempt to solve all of the cloud native challenges and integrate all of the platform tools into Kubernetes. However, Tim highlighted that this had been tried before with OpenStack… and with debatable results.

This concept was nicely summarized in a tweet from Tim Bannister that Tim shared on the big screen: “Kubernetes should stay unfinished.” To quote one of my favourite comic/movie franchises, Kubernetes may be “the hero Gotham needs, but not the one it deserves right now”.

Don’t forget about developer experience!

It was great to see several mentions of developer experience and the importance highlighted in the KubeCon NA keynotes. KubeCon has historically been an infrastructure and operations-focused event, but we mustn’t forget about our primary (platform) customers: the developers.

My buddy Kasper Nisson was on a roll with his tweets related to developer experience, capturing the keynote panel:

And also valuable thoughts from the Backstage Con co-located event:

I also heard a lot of good things about the AppDeveloperCon co-located event. Dapr (covered later) was mentioned a lot in conversations about this event, alongside TestContainers for testing, Dagger for CI/CD, and the benefits of ephemeral clusters and PR-based preview environments (courtesy of the Okteto team). I also bumped into mentions of the devcontainer spec a few times, alongside the concept of cloud development environments, frequently with a tip of the hat to Loft Labs’ DevPods and GitPod’s CDEs.

As a shameless plug, I’m helping several clients in the cloud native dev tooling space with product marketing and go-to-market strategy at the moment, and focusing on developer experience and “jobs to be done” is a big part of this work.

You can’t successfully position your product if you don’t know who your users are and their most significant pain points. Developers are an intelligent bunch who like to solve their own problems, and often, your product is going head-to-head with the status quo solutions that use duct tape and bailer wire but get the job done “good enough”.

Increased focus on app development and integration

I’ve long been impressed by the work the Dapr community are doing, and they are arguably leading the way in highlighting the application developers’ needs within the cloud native community. For folks not familiar with the CNCF project, Dapr “provides integrated APIs for communication, state, and workflow.” For those of us with a few grey hairs, think of it as enterprise middleware for the cloud generation (done well).

Even with the rapid adoption of AI/LLMs in the cloud native space, we still need to focus on defining the correct abstractions; otherwise, we could generate a lot of incomprehensible code!

In my opinion, Dapr provides several very useful abstractions required for building distributed apps, such as pub-sub, state management, distributed locking, etc.

I attended several of the Dapr talks and was impressed. If I were building a greenfield cloud native application (or even evolving a brownfield app), I would look at this project first:

The open source community is strong, and the folks at Diagrid offer several great commercial solutions that reduce the friction and operational burden of managing your backing infrastructure. I had several great chats with Mark Fussell and Bilgin Ibryam, and I encourage you to follow them on X. My InfoQ buddy, Thomas Betts, also wrote a summary of the latest Diagrid commercial release: “Diagrid Launches Catalyst, a Serverless, Fully-Managed Dapr Offering.”

Cloud native communications: Bundling FTW?

As Jim Barksdale, former CEO of Netscape, once famously stated, there are “only two ways to make money in business: one is to bundle; the other is unbundle”. And we’re seeing a lot of bundling in the cloud native ecosystem!

One domain in which this observation really stood out is the cloud native communication stack.

I’ve discussed this several times, including a recent presentation: “The Busy Platform Engineers Guide to API Gateways”. The vast majority of platform builders will need three components in this stack: API gateway (for north-south traffic management), service mesh (for east-west traffic), and container-native interface or CNI (for all of your lower-level software-defined networking).

I’ll own my bias here — having worked for Ambassador Labs, maker of Edge Stack, and being a big supporter of the Linkerd folks — I like picking “best in class” solutions and integrating these.

However, the commercial winds are tending towards bundling. I saw various vendors in this space moving up and down the stack with their latest offerings, e.g. Isovalent (creators of Cilium CNI) are pushing up the stack by embracing service mesh and API gateway. Solo.io are pushing down the stack by incorporating CNI (Cilium) into its Istio service mesh and API gateway offerings. Tetrate, Kong, Tyk, Traefik, and others are making similar plays.

If you follow the cycle (and the money), the smart play appears to lean towards bundling.

And I believe this learning doesn’t just apply to the domain of cloud native comms. In general, enterprises are looking to consolidate vendors, reduce the total cost of ownership (TCO) of DIY solutions, and shift cloud spending commits onto end-to-end solutions offered within cloud marketplaces.

On a related note, the service mesh battle panel, moderated by Keith Mattix, was one of my favourite sessions at the conference. The participants weren’t holding back, leading to a fun and informative discussion versus the usual format of everyone agreeing!

Security is big business

I’ll keep the final three takeaways brief, but I wanted to mention that the vendor hall was full of security solutions, from secure supply chain solutions (SBOMs and SLSA) to network intrusion detection and permission management (OPA). There is a lot of interest (and money) in this space.

The sessions also provided good coverage of the topics, and KubeCon co-chair Frederick Kautz and the CNCF folks were keen to emphasise the importance of security (and related CNCF projects) in the keynotes:

More focus on sustainability: Observability, scaling & FinOps

Also of prominence in the keynotes was the topic of sustainability. There was an interesting panel session, “Environmental Sustainability in the Cloud Is Not a Mythical Creature”, and even the AWS-sponsored keynote focused on this theme:

This is an important topic, and it was great to see the underlying drivers of understanding and monitoring your cloud (carbon) footprint being explored via observability, automated scaling, and FinOps.

For those new to FinOps, I learned a lot from Roi Ravhon in this recent InfoQ podcast I recorded with him, “Roi Ravhon on FinOps, Application Unit Economics, and Cloud Cost Optimization.”

Community, community, community

I can’t write a KubeCon summary without mentioning the power of the community. Seeing so many success stories and awards in the keynotes was awesome.

One of the primary reasons I attend these events is to catch up and hang out with folks in the community. The road to cloud native is not always easy, and I enjoy swapping stories with fellow travellers. I was a bit lazy with photos this year, but my devrel buddies Marino Wijay and Daniel Oh had me covered:

And I also got to spend some quality time with my old Ambassador Labs teammates:

I didn’t get a chance to catch up with everyone I wanted to (KubeCon is busy!), so feel free to contact me via socials if you want to arrange a chat via Zoom. Otherwise, I hope we meet soon on the conference circuit!

Wrapping up

As I mentioned at the top of this article, I had a blast attending this KubeCon. The community and surrounding support systems are clearly maturing, and now we have the tricky task of balancing standardization and innovation.

I’m bullish that the cloud native future looks bright. I do, however, think we’ll see more shades of AI/LLMs in the future 🙂

If you work in developer relations or technical go-to-marketing, be sure to subscribe to my Avocado Bytes newsletter, and I’ll see you in Paris for KubeCon EU 2024 (don’t forget the CFP closes on the 26th November)!

--

--

Daniel Bryant

Platform Engineering @Syntasso | News/Podcasts @InfoQ | Web 1.0/2.0 coder, platform engineer, Java Champion, CS PhD | cloud, K8s, APIs, IPAs | learner/teacher