Profiling has been a part of the software engineering toolbox since the 1970s, but it was long regarded to be too high in overhead, so it was only done ad hoc. The problem with profiling that way is that you never catch the moment in time where for example an incident happened. Continuous profiling is the systematic capturing of profiling at all times, Google has been doing it for over a decade, but it comes both with challenges of how to capture the data at low overhead and also storing and querying the enormous amount of data that continuous profiling brings with it. Polar Signals created the open-source project Parca to bring continuous profiling to the masses. Frederic Braczyk, Founder and CEO of Polar Signals joins the show to discuss the uses of continuous profiling as well as the challenges in building everything from the low overhead profilers using eBPF to purpose-built storage engines for the high volume data it entails.
Sponsorship inquiries: sponsor@softwareengineeringdaily.com
The post Continuous Profiling using eBPF with Frederic Branczyk appeared first on Software Engineering Daily.