Conference Review
KubeCon NA 2022 Summary: Maintainers, Open Standards, and the Rumoured Demise of DevOps
The future of Kubernetes is platform-shaped: self-service, observable and secure, and driven by the community
The Ambassador Labs team and I are still buzzing from another amazing KubeCon in person. Building on the success of KubeCon EU in Valencia, the NA event in Detroit was almost back to the full pre-Covid experience. It was amazing to connect with so many folks again, and have great conversations with friends, old and new.
I also had a blast presenting with Flynn from the Buoyant team, “Emissary + Linkerd Resilience Patterns: Rate Limits, Retries & Timeouts” (there’s nothing like presenting on a real stage!), and Alice and Flynn also delivered a great Emissary-ingress maintainers session. Many thanks for all of the great feedback!
So, what did we learn from KubeCon NA 2022? These are my top 10 takeaways:
- Building for the road ahead: The importance of community and maintainers
- The future is built on open standards
- Developer experience is now table stakes
- Developers are struggling with fast feedback
- DevOps is not dead… but platform engineering sure is popular
- The search for platform orchestration (and abstraction)
- Wasm appears to be the next hottest runtime
- Hot tech: eBPF and secure supply chain/SBOMs
- Mo’ Money, Mo’ Problems: Observability, security, and AIOps
- Lift, shift, and modernize: Legacy tech is so last year
We would like to say a big thanks to the CNCF for organizing the event, and also a huge thank you to all the speakers, sponsors, volunteers, and attendees. You all rock!
If you’re looking for summaries of the keynotes, check out these threads by the always-awesome Edidiong Asikpo!
And don’t forget our “K8s in 8” learning challenge and competition is still running (riffing on our popular “8 Mile” theme at the event :) )
Building for the road ahead: The importance of community and maintainers
From the moment we walked into the venue, the “building for the road ahead” signage was on full display. With over 170k+ contributors to the cloud native ecosystem, the opening keynote leaned heavily into the benefits and service that project maintainers provide. Priyanka Sharma, executive director at the CNCF, explored the responsibility we all have to contribute back to the ecosystem and community — both as individuals and also as organizations.
Also discussed in the opening keynote was Gartner’s prediction that cloud native platforms will serve as the foundation for 95% of new digital initiatives by 2025. The concept of a “platform”, its constituent parts, and how these will be integrated were strong themes of the entire conference.
The future is built on open standards
In a great Friday keynote, Frederick Kautz, Co-chair of KubeCon, (and wearing a unicorn custom, no less!), shared his future of the cloud native ecosystem, and this leaned heavily on integration and the need for open standards:
I saw this echoed throughout the event. As I work in the ingress and networking space, I had lots of great related discussions at the event, with folks frequently asking about the Gateway API and Service Mesh Interface (SMI).
You also couldn’t walk the length of the sponsor hall without hearing good things about Open Policy Agent (OPA) and OpenTelemetry in the observability space and SPDX in the secure supply chain and SBOM space.
Developer experience is now table stakes
This is the first KubeCon in which attendees I was chatting to at the booth were freely using the term “developer experience” without air quotes or irony.
It appeared second nature for attendees to be asking what our developer experience was like (even with traditionally ops-focused products like Emissary-ingress and Ambassador Edge Stack), and I also noticed the same ease of usage of the term when people were talking about both tools they liked and also their own development and continuous delivery toolchain.
I’ve been banging on about the cloud native developer experience for a number of years, and now I believe that the developer experience penny has finally dropped. Let’s hope this leads to increased focus (and spending) on improving the UX for the tools we all use day in and day out.
Developers are struggling with fast feedback
I was talking to a lot of folks about Telepresence and related tooling (with a big shout to Ricardo Rocha, a computing engineer at CERN, for mentioning this CNCF tool in his Thursday keynote!) A lot of developers were talking about the long build, push, and test cycles they were enduring, with roughly half not realizing there are solutions available and the other half simply thinking this is the way it is with cloud native development (“hey, there’s always a tax when adopting new technologies”).
If you haven’t played with Telepresence, the simple pitch is that it “puts your laptop in your remote Kubernetes cluster” via a two-way networking proxy. This means you can interact with remote services as if they were local e.g. curl my-service.my-namespace:port/path and also route remote traffic to your local machine for testing your services running in your local IDE, debugger, or profiler etc.
Telepresence plays nicely with all of your local toolchains, and there is also a Docker Desktop Extension. If you want to know more, Felipe Cruz from Docker and I will be presenting a webinar on November 8th, “Building Microservice Systems Without Cooking Your Laptop”
DevOps is not dead… but platform engineering sure is popular
If you were following the Twitter #KubeCon hashtag, you couldn’t have failed to notice some intense discussion around the “DevOps is dead” concept being promoted by some folks at the event. There was a general consensus that DevOps is not dead, but new sub-themes are emerging, such as platform engineering (following on from site reliability engineering).
(source: https://www.linkedin.com/posts/charity-majors_devops-iac-ansible-activity-6986492951031947264-zJs1/)
In fairness to the main perpetrators of the “DevOps is dead” statement, they were pitching the same ideas of platform engineering being an evolution of DevOps when I took the time to have a chat with them.
The big focuses in this space are: creating and using the correct abstractions, supporting a more effective developer experience, also enabling developer self-service:
The search for platform orchestration (and abstraction)
Related to the above key takeaways, there were several companies in the sponsor showcase that were offering platform orchestration tooling. The target audience was DevOps teams and platform engineers, and the products were often integrated within internal development platforms (IDPs) or developer control plane (DCP) offerings.
I had several great chats with folks about this, and the conclusion that we all appear to be gravitating to is that there won’t be a “Heroku for Kubernetes.” Instead, there will be a series of “build your own Heroku for Kubernetes” components, which allow you to assemble your own platform product (or platform as a product).
Frameworks like Crossplane are leading the charge here in relation to platform abstractions (and received multiple mentions in the keynotes and sessions), and I’m also keeping an eye on the Open Application Model (OAM) and Kratix from the Syntasso folks. The Backstage project also featured heavily in related discussions around implementing IDPs and developer portals, and there was even a dedicated co-located event, BackstageCon.
Wasm appears to be the next hottest runtime
The week began with lots of buzz around Docker announcing Wasm development support in technical preview at the Cloud Native Wasm Day. Fermyon, who offer a Wasm development toolchain and cloud platform, also announced they had raised a $20M Series A round of funding, and the Cosmonic folks were showing off a PaaS for Wasm.
Although we all love containers, the Wasm story is compelling, particularly if you are building for “edge” or “serverless” uses cases where a low resource footprint, quick startup time, and sandboxed environment are requirements.
Hot tech: eBPF and secure supply chain/SBOMs
In addition to the buzz around Wasm, there were a lot of good things being said about eBPF. I bumped into Liz Rice on the final day of the event, and her arm was still tired from signing all the copies of her “What is eBPF” O’Reilly report at the Isovalent booth :) Seriously though, I’ve learned a lot of good things about eBPF from the Cilium service mesh use case and also from Project Falco’s Kubernetes threat detection framework and New Relic’s Pixie observability tooling.
In the interest of impartiality, there was also a bunch of interesting work going on with other service meshes, including Ambient Mesh with Istio, and the continued great work (and focus on simplicity and developer experience) with Linkerd by my friends at Buoyant.
Changing gears, but there was a lot of focus on secure supply chains and SBOMS at KubeCon, both in the sessions and also in the sponsor showcase. Standards are still emerging here, and so you may have to attempt to pick a winner, but the most important thing is that developers are fully understanding the need to think about this early in the SDLC (if you’re interested in this space, check out this Ambassador Labs podcast I recorded with Kelsey Hightower).
Mo’ money, Mo’ problems: Observability, security, and AIOps
Riffing on the topics mentioned in the section above, it was obvious from looking at both the session program and sponsor showcase that there were three trends being heavily monetized, each with varying degrees of success: observability, security, and AIOps.
The first two have well-established use cases, and may even be recession-proof to some degree (you always have to know if your systems are performing as expected and are secure, right?), the third use case feels like a riskier bet at the moment. My colleague, Dave Sudia, and several people I chatted with at the event (mostly ops folks in this context) were unsure about trusting ML for automating operations or incident management, particularly if this unproven tech was going to cost a lot.
Lift, shift, and modernize: Legacy tech is so last year
A final topic that I wanted to highlight comes from many great discussions at the Ambassador Labs booth. Attendees appeared to no longer be satisfied with simply “lifting and shifting” applications into the cloud. They wanted to go completely “cloud native”. This often meant adopting Kubernetes with the goal of taking full advantage of all of the benefits that cloud computing has to offer: using the latest innovations “as a service” (e.g., AI/ML, columnar databases, etc.), fully elastic compute, less operational overhead, and increased development velocity.
We heard from a number of large organizations looking to move away from older proxy technologies like OpenResty to Envoy, primarily for performance reasons, but also because configuring and running these tools wasn’t done with the same approach as other tooling was standardizing on e.g. using custom resources (and GitOps) for configuration, and relying on Kubernetes to provide resilience, manage state, and scale effectively.
Until next time in Amsterdam… or Chicago!
And so that’s a wrap! I’ll again say a big thank you to all that were involved with the conference. It’s great to be back in person, and I look forward to connecting again at the next KubeCon EU in Amsterdam or KubeCon NA in Chicago.
I’ll leave you with one of the most picturesque moments from the event:
(Don’t forget to check out our “K8s in 8” learning challenge and prizes!)
Thanks to Dave Sudia, Richard Li, and Mark Trang for providing feedback on this article!
