AWS re:Invent 2024 - How to build scalable platforms with Amazon EKS

Have you ever built a shiny new platform for your developers, only to find that no one wants to use it? At AWS re:Invent 2024, Nirmal Mehta (Principal Solutions Architect, AWS), Isaac Mosquera (Head of GTM for Containers & Serverless, AWS), and John Weber (Senior Director, Developer Services, Adobe) talked about this common problem and shared ways to solve it. They showed how to build platforms that developers actually want to use, focusing on trust, scalability, and the power of Amazon Elastic Kubernetes Service (Amazon EKS).

Understanding the Trust Challenge

Nirmal began by explaining why many platforms fail to attract users. He introduced the concept of trust in platform engineering, crucial for a platform's success. Trust is built on three main things:

• Logic: Whether the platform makes sense
• Authenticity: If there's true collaboration with developers
• Empathy: Caring about what developers need.

Next came the "Field of Dreams" fallacy, where platform teams think that just building a feature-rich platform will automatically attract developers. This "If you build it, they will come" approach usually fails because it ignores the importance of trust. Lastly, Nirmal addressed the challenge of balancing growth and trust. As platforms get bigger, it becomes difficult to meet the needs of more users while maintaining trust. The key is figuring out how to scale the platform without eroding the trust that's been built

Rethinking Automation and Scale

Isaac Mosquera then delved into why traditional automation approaches often don't work well:

1. The automation cycle: Many organizations find themselves in a yearly cycle of trying to automate more, yet never fully achieving their goals. This is often due to increasing complexity as systems grow.
2. The problem with pipelines: As organizations scale, they tend to break down into smaller teams, each with their own tools and processes. This leads to complex pipelines and communication challenges.
3. A new approach - API-driven platforms: Instead of focusing on pipelines, Mosquera proposed thinking in terms of APIs. This new way of thinking can help teams talk to each other better and make things run on their own, instead of getting stuck in lots of meetings and paperwork.

Leveraging Kubernetes as a Platform Framework

Mosquera introduced a novel way of using Kubernetes:

1. Beyond container orchestration: While many see Kubernetes as just a tool for managing containers, it can serve as a complete platform framework.
2. Key benefits of Kubernetes as a platform: Kubernetes offers several advantages when used as a platform framework:
   • Built-in API handling and Role-Based Access Control (RBAC)
   • Extensibility through custom controllers
   • A single state store (etcd) for all platform components
   • A large ecosystem of open-source controllers
3. Amazon EKS for scalable control planes: With this approach putting more pressure on the Kubernetes control plane, Amazon EKS provides a scalable, managed solution that allows teams to focus on building their platform rather than managing infrastructure.

Real-World Implementation: Adobe's Ethos

John Weber shared how Adobe put these principles into practice with their internal platform, Ethos. Ethos is Adobe's company-wide platform that helps developers build, deploy, and manage their applications more efficiently. Here's how they built and improved Ethos:

1. Treating developers as customers: Adobe focused on understanding and meeting the needs of their developers, just as they would for external customers.
2. Building a community: They started a program to get people excited about the platform and let anyone in the company help improve it.
3. Creating a culture of accountability: Teams were made responsible for their own code, including on-call duties and incident response.
4. Implementing cost transparency: Adobe built granular dashboards and cost attribution tools to help teams understand and optimize their resource usage.
5. Using service level indicators (SLIs) and objectives (SLOs): This helped set clear targets and measure platform performance consistently.
6. Leveraging Amazon EKS: By using EKS, Adobe improved their cluster-to-operator ratio from 10:1 to 30:1, significantly reducing operational overhead.

The results were impressive: developers can now deploy a full greenfield stack in just a few days instead of a month. Also, six times more people from other teams are now helping to improve Ethos

Building Attractive Platforms: Key Takeaways

The speakers wrapped up with some key advice for building platforms that developers will want to use. They stressed the importance of building trust by being open, working together, and making sure the platform works reliably. They suggested treating the platform like a product that's always getting better, listening to what users say and making changes based on their feedback.

Using Kubernetes and its tools can help hide complicated tech stuff from developers. They also recommended working with people outside your company, like open-source groups and cloud experts. The speakers talked about CNOE (Cloud Native Operations Excellence), which is a free set of tools and ideas for building cloud platforms. The main things to remember are focusing on trust, using open standards, working with others, and always improving based on what developers need.

For those who want to learn more, the speakers shared a GitHub repository link with resources to help build platforms faster. The full session recording is available on the AWS YouTube channel, featuring detailed explanations and demonstrations from Nirmal Mehta, Isaac Mosquera, and John Weber.