Overview
Explore continuous profiling in cloud-native environments through this conference talk. Discover how techniques from Google's "Google-Wide Profiling" paper can significantly reduce fleet-wide resource usage. Learn about ad-hoc profiling for CPU and memory analysis, and see how continuous profiling enables new workflows for systematic profile collection. Gain insights into profiling with Go and witness a demonstration of Conprof, an open-source continuous profiling project. Understand how continuous profiling provides a comprehensive fleet-wide understanding of code at runtime, guiding the development of robust, reliable, and performant software while reducing cloud expenses. Delve into topics such as pprof for various languages, Go and JS examples, the Parca profiler overview, compression techniques, storage solutions, and the creation of metastore entries and profile trees. Explore the roadmap for Parca and discover how to implement continuous profiling in Go code to optimize software performance and efficiency.
Syllabus
Intro
About us
Profiling: Why?
pprof for other languages
Code
Go & JS example
The problem
Continuous Profiling
High level overview of Parca profiler
Previously: Compression
Previously: Improved Compression
Parca's Storage
Creating a Metastore Entries
Creating a Profile Tree
Append Profile Meta
Append Profile Tree - Sparseness
Combine/Merge Profiles
Parca's Roadmap
Taught by
CNCF [Cloud Native Computing Foundation]