Скачать или смотреть The Curious Case of the Latency Spike

Presented at o11ycon+hnycon, June 9–10, 2021
https://www.honeycomb.io/

Michael Ericksen, Staff Site Reliability Engineer @ IMO

The Curious Case of the Latency Spike

This talk investigates the aftermath of a production issue where performance diverges significantly from expected boundaries. To do that, the talk follows an application team as they debug and instrument to understand why their application behaves significantly worse under certain load parameters.

In this talk, Michael covers false starts during early issue investigation including cold starts, provisioned concurrency, and function memory limits. You'll learn how and why increasing application observability enabled my team to bypass invalid assumptions about root causes of poor application performance, how we moved from guessing to knowing, and how that focus helped us fix application behaviors that most directly impacted our customer experience. This talk also explores how observability sharpen's a team's ability to identify the right resource investments to make and how SLIs and SLOs provide the clarity teams need to prioritize and assess production issues.

If you see a problem with captions, email [email protected].