Published
11/23/21
written-by
share

In this episode of GV News, Polar Signals CEO and founder Frederic Branczyk joins Google's Kelsey Hightower and me to discuss the evolution of infrastructure management, and the increasing importance of continuous profiling as an observability tool for modern developer operations (DevOps) organizations.

I first met both Kelsey and Frederic when they worked at CoreOS, a beloved GV portfolio company pioneering a next-generation Kubernetes orchestration platform well before the world had fully discovered the power of Kubernetes. Before joining forces with the publicly-traded, open-source company RedHat, the CoreOS team contributed groundbreaking work in containers, monitoring, and infrastructure management. Kelsey went on to become a beloved voice of developers and customer empathy in enterprise software. Frederic founded Polar Signals to tackle another big hurdle — system observability — with a passion for enabling DevOps teams to gain visibility into the complex systems they manage.

"Think of your car," explains Kelsey, characteristically simplifying engineering-speak into a straightforward analogy, as he tends to do so well. "Imagine driving a car with no dashboard, no gas meter, no maps. You have no idea of what's going on until you're out of gas and calling a tow truck."

Older monitoring approaches check system health periodically, not constantly. As Frederic notes, "It turns out that software often goes wrong when we're not looking. Continuous profiling means that in the background, we're still collecting data and insights about system health. We can understand, in real time, which components are using CPU time and memory. We understand the implications of that data. Simply put, if you don't measure, you don't know."

Polar Signals is rooted in open source tech, which GV has a deep track record of supporting, including star open source companies like Cockroach, GitLab, Snorkel, and Vectorized. Recently Frederic and his team announced Parca, its open source project for continuous profiling.

"A lot of what went into Parca was inspired by my work with Prometheus in the past," notes Frederic. "The agent uses eBPF to collect this data at even lower profiling, which lowers the overhead by orders of magnitude. I built this for myself, and then I realized that everyone needs this kind of software. Think of the value for just cost optimization. Who doesn't want to save money on their cloud bills?"

The last four companies I've funded have been remote-only, and Polar Signals is no exception. In the middle of the pandemic, Frederic founded the company with a globally distributed team, saying "We're building to be a remote company forever."

Along with the security industry, Google was one of the first champions of continuous profiling, as a way to get a more accurate picture of what's really happening across a system and broader, across a whole datacenter. Now, early-stage companies like Polar Signals, says Kelsey, "are really leading the charge here. It's easy to predict the future when you're working on it. This is really ahead of the curve."