Metrics and traces provide important context for understanding what an application is doing and finding areas for improvement. But sometimes there are lower-level issues that emerge from how code interacts with runtime environments, or crashes and constraints that result from problems with garbage collection and memory management. When standard observability practices don’t provide enough insight, how do you know where a problem is coming from and what to optimize?

In this talk, Zach McCormick, Senior Staff Engineer, will discuss how Braze uses Continuous Profiler with their Ruby application to identify and fix CPU, network, and memory issues in their platform. He'll go over the symptoms of these problems, how they used profiling to confirm them, and the strategies employed to boost overall system performance, resulting in increased efficiency and reduced resource costs. Attendees will learn how profiling fits into an overall observability approach and a problem-solving framework for identifying and solving low-level problems.