AWS re:Invent 2025 - Why is Reliability So Hard? (DVT227)

Summary of AWS re:Invent 2025 Presentation: "Why is Reliability So Hard?"

Introduction

• Presenter: Hannes Lank, CEO and co-founder of Czechly
• Czechly helps organizations detect, communicate, and resolve software reliability issues faster
• Goal is to help engineers "own reliability from pull request to postmodern software"

The Evolution of Software Reliability

• A decade ago, software was built and shipped much less frequently (yearly, quarterly, monthly)
• Today, software is built and shipped almost instantly, but reliability has not kept pace
• Yesterday's applications were simple, with few dependencies - issues were easy to identify
• Modern applications are highly complex, with many dependencies that introduce potential failure points

The Reliability Challenge

• Increased complexity and dependencies make it harder to ensure reliability
• Traditional approaches of more people, processes, and testing have not solved the problem
• Both development and operations teams try to validate application functionality, but in siloed ways

Key Principles for High-Performing Teams

1. Predictability: Ability to predict how an application will behave when released to production
2. Accountability: Knowing what changed, who changed it, why, and when - to identify the root cause of issues
3. Resiliency: Building applications that can be quickly rolled back in the event of problems

Czechly's Approach

• Unifies testing and monitoring into a single, version-controlled workflow
• Allows teams to build tests (UI, API, uptime) as code and deploy them for continuous monitoring
• Integrates the reliability pipeline with the CI/CD pipeline, enabling a common language and visibility

The Evolving "You" in Software Reliability

• Traditionally, "you" referred to the developer or engineer responsible for the code
• Today, "you" encompasses anyone who touches the user experience, including agents, cloud code, and other tools
• In the future, agents may be capable of building, testing, monitoring, and owning more of the software lifecycle

Key Takeaways

• Reliability has not kept pace with the rapid evolution of software development
• Increased complexity and dependencies make it harder to ensure reliability using traditional approaches
• High-performing teams focus on predictability, accountability, and resiliency to improve reliability
• Czechly's approach unifies testing and monitoring, integrating the reliability pipeline with CI/CD
• The concept of "you" in software reliability is expanding to include a wider range of stakeholders and tools