In order to truly optimize our code we need to be able to observe its behavior in production. Once we identify inefficiencies we need to drill further down to find the actual bottlenecks - the functions and memory allocations that are costly and slow. And get that information back to the developer's desktop for continuous optimization.
Exactly for that high granularity analysis we need the practice of continuous profiling with periodical sampling of code execution and making it accessible to engineers and usable for Profile Guided Optimization.
In the talk we'll go over the leading code profiling tools, practices and formats and see how to implement continuous profiling in a Kubernetes cluster.