Keynote
The Cloud Native Feedback Loop
It's the feedback loop of an end user to a project that can be the breakthrough of a new project. Use Reference Architectures to see what other companies have built. Projects go from Sandbox, to Incubating to Graduated. Graduated projects are here to stay. The fourth stage, that we don't really talk about, is Archival. It's important that projects take their lessons learned and start something new sometimes.
Use LFX Insights to find out more on a project's health score.
HAProxy
We all assumed that our infrastructure would be simple and easy to manage. The business evolves, so the IT has to grow with it. You have to innovate to grow within your market.
The proven solutions to connect systems like routers, firewalls and VPNs are not very flexible. On the other hand, we have the service mesh. The point is to make the application traffic more observable, but only works a bit in a fractured environment. As an industry, we need to come up with some universal solution to connect anything together. Universal Mesh wants to connect all existing solutions in your datacenters.
Orchestrating Games
Ubisoft wants to use open software to avoid cloud vendor proprietary features. CNCF projects can also evolve faster than any software anywhere developed in-house. Instead of playing the catchup game with providers, Ubisoft leverages CNCF.
If players are plaing a game, you cannot shut down the server. Agones aims to resolve the issue that this does not really scale well. One of the recent examples, is Rainbow Six Mobile. Diffent regions, but the infrastructure can adapt to whatever cloud provider is best at the given moment.
Cloud Native Platforms
There is always one bottleneck that is holding back your platform. That bottleneck should be your most important issue to solve. But there will always be a bottleneck in the platform.
The biggest bottleneck in platform engineering is growing capabilities. One of the benefits of microservices is that we unblocked developers to work in parallel. AI has changed the bottleneck. There's a bunch of new features, and we can produce garbage code at a record speed.
Event when using AI, Agents can only move as fast as their platform. Don't bottleneck the AI, give it access to the cloud. That's not innovation, that's shadow IT. Platform capabilities are for life. We need to build capabilities with long term support in mind.
Vertical scaling in platform engineering would mean adding developers to the team. Great for job security, bad for everything else. What does horizontal scaling look like? Allow users to work indepedently. We need a self-service capability creation API so users can both supply and consume capabilities of the platform.
A decade ago, Heroku defined the https://12factor.net/. This framework gives guardrails on how a platform can define good consumer citizens. The platform still needs to define a good producer citizen. https://github.com/Cloud-Native-Platform-Engineering/cnpe-community is the first iteration of defining good platform producers.
Kubernetes for Accelerated Workloads
For developers, it's all about velocity. Kubernetes can provde velocity and scaling without much intervention. Platform teams can provide efficiency, but the ecosystem is where it really happens.
In the journey of making Kubernetes the OS layer, the fist step with DRA is already semi-finished. The next step with Workload Aware Scheduling will optimize this further. Please let this be the end of Airflow.
DX to AX
The next decade of Netlify will see a shift from Developer Experience to Agent Experience. Builders are pushing out frontend sites with a gradient at record pace. Netlify built the infrastructure already to integrate well with Agents.