Title

AWS re:Invent 2022 - Optimizing price performance with feature management and observability (PRT259)

Summary

Brian Rinaldi from LaunchDarkly and Charity Majors from Honeycomb discuss the importance and benefits of feature flags in application development.
Feature flags are more than simple Booleans; they enable safer and faster deployment and release processes by separating the two.
LaunchDarkly offers a feature flagging service with high performance, cross-platform SDKs, global flag delivery network, advanced features, and integrations with existing tools.
Honeycomb uses feature flags in conjunction with observability to manage deployments, reduce risks, and handle backend changes effectively.
Observability is defined as instrumenting code for high cardinality and dimensionality events, allowing for detailed inspection and analysis of software behavior.
The speakers emphasize the moral and technical responsibility to decouple deploys from releases, highlighting the importance of feature flags and observability in modern software delivery.
Real-world examples from Honeycomb demonstrate how feature flags and observability can lead to significant cost savings, improved user experience, and more controlled and intentional changes in production.

Feature flags are a critical tool for modern software development, enabling teams to deploy code to production multiple times a day without breaking things.
The separation of deploy and release processes allows for progressive rollouts, targeted releases, and the ability to quickly turn off features that aren't working, without the need for rollbacks or redeployments.
LaunchDarkly's feature flagging service is designed for high performance and ease of use, with a focus on scalability and integration with other tools, such as Honeycomb for observability.
Observability, as defined by Charity Majors, is not just monitoring or logging but involves detailed instrumentation of code to ask and answer complex questions about the software's internal state.
The combination of feature flags and observability is presented as essential for modern software delivery pipelines, allowing for safer, faster, and more confident deployments.
The talk provides practical examples of how Honeycomb uses feature flags and observability to manage frontend changes, optimize backend processes, and experiment with new technologies like AWS Graviton.
The speakers advocate for a cultural shift in engineering practices, where speed and frequent deployments are seen as safety measures rather than risks, and where engineers take ownership of their code in production.