Observability in software systems has undergone significant transformation, and one of the key players driving this change is Polar Signals. Founded by Frederic Branczyk, the company has established itself at the forefront of continuous profiling and eBPF-driven observability. In this episode of Let’s Talk, Branczyk shares his journey, the inspiration behind Polar Signals, and the future of observability.

Branczyk also discusses a case study where eBPF helped Polar Signals reduce their Google Cloud bill by measuring and optimizing cross-zone traffic. He emphasizes the importance of continued innovation and collaboration, highlighting how eBPF will play a crucial role in shaping next-generation observability solutions.

Polar Signals: From observability to profiling

Branczyk’s journey into observability started over a decade ago, where he made significant contributions to the Prometheus project, including the Prometheus Operator and the Kube-Prometheus-Stack. His work at CoreOS, which was later acquired by Red Hat, further solidified his expertise in Kubernetes observability. While leading a team of engineers at Red Hat, he encountered a key challenge: optimizing resource usage in performance-sensitive environments.

A turning point came when he came across a Google paper discussing the benefits of continuous profiling—constantly monitoring resource usage across all running processes. Seeing the potential, Branczyk realized that profiling could be made more systematic and accessible, similar to how Prometheus transformed time-series monitoring. This revelation led to the founding of Polar Signals in late 2020.

Why eBPF? 

eBPF (Extended Berkeley Packet Filter) is the cornerstone of Polar Signals’ profiling technology. One of the key advantages of eBPF is its efficiency. Branczyk explains that the entire profiling tool developed by Polar Signals was just 90 lines of code, with most of it dedicated to struct definitions for communication between kernel and user space. Unlike traditional monitoring systems that generate excessive data, eBPF enables targeted data extraction with minimal complexity.

eBPF Case Study: A recent case study highlights how Polar Signals leveraged eBPF to solve an internal challenge: reducing cross-zone traffic costs in Google Cloud. By utilizing eBPF with netfilter, they built a tool that provided visibility into inter-zone traffic at the pod level, enabling them to optimize workloads and significantly cut cloud expenses. This project, initially developed for internal use, has now been made available for others facing similar challenges.

Branczyk advises those exploring eBPF to start with existing libraries and tools such as libbpf for C/C++, libbpf-rs for Rust, and Cilium’s Go-based libraries. Additionally, tools like bpftrace provide a lightweight way to experiment with eBPF before diving into more complex implementations.

The future of observability: Unifying metrics, logs, and tracing

Looking ahead, Branczyk envisions a shift toward a more integrated observability ecosystem. Currently, observability is fragmented across logs, metrics, tracing, and profiling. However, emerging databases designed specifically for observability are beginning to consolidate these data types into a unified platform. Companies like Datadog (with their acquisition of Quickwit) and Polar Signals, which has developed the columnar database FrostDB—are pushing this vision forward. As observability continues to evolve, eBPF will remain a key enabler of efficient, real-time monitoring.

Guest: Frederic Branczyk
Company: Polar Signals
Show: Let’s Talk

This summary was written by Monika Chauhan.