Polar Signals open-sources Parca to optimize code and cut cloud bills

Let the OSS Enterprise newsletter guide your open source journey! Sign up here.

“Continuous profiling” might not be a familiar concept to every company, but in a world increasingly drawn to cloud software and infrastructure, it’s something that most should be aware of. It’s a signal of sorts that belongs to the broader software monitoring category known as observability, which is concerned with measuring the internal state of a system by analyzing the data outputs — this can tell companies how their software is performing, and identify issues.

Continuous profiling, specifically, is all about monitoring the resources that an application is using, such as CPU or memory, giving engineers deeper insights into what code — down to the line number — is consuming the most resources. A common use case is in helping companies reduce their cloud bill, given that most of the major cloud platform providers charge on a consumption basis: the more consumption, the higher the cost.

So continuous profiling is basically optimizing codebases to save on cloud costs. Google was one of the early champions of the practice, detailing it in a 2010 white paper titled, Google-Wide Profiling: A Continuous Profiling Infrastructure for Data Centers.

There are several notable players in the space, such as software monitoring giant Datadog, while Andreessen Horowitz-backed Optimyze, which develops the closed source Prodfiler, does something similar.  And newcomer Polar Signals has officially thrown its hat into the ring today with the launch of a new continuous profiling open source project called Parca, which is available on GitHub now.

Additionally, Polar Signals today announced it has raised $4 million in seed funding from Alphabet’s venture capital arm GV and Lightspeed.

The backbone of Polar Signals’ open source project

Founded in 2020 by Frederic Branczyk, a former Red Hat senior principal engineer and prominent figure in the Prometheus and Kubernetes open-source ecosystems, Polar Signals is designed for large-scale infrastructure, which means it’s gunning for the enterprise segment in a sizable way.

Parca is the backbone of Polar Signals, and as an open source project, it’s designed to bring the power of continuous profiling to developers from all businesses. It packs a bunch of features out of the box, including capabilities for collecting, storing, and making profiles available for query over time — this includes CPU profiling to determine the amount of time a CPU needs to execute a specific piece of code.

Above: Parca: Point-in-time CPU metrics

Polar Signals has been designed from the get-go to play nicely with all the usual observability tools, such as Jaeger and Prometheus, the latter now being the “defacto standard” for monitoring any Kubernetes environment. “We have taken special care that Parca and Polar Signals integrate particularly well with those environments,” Branczyk told VentureBeat.

The Parca agent is deployed into each Kubernetes cluster node, with the workloads automatically profiled with “super-low overhead,” Branczyk added. “We have prepared lots of pre-baked deployment options and tutorials to make this as easy as possible — users can then choose to run the storage themselves, or purchase the hosted version from us.”

The commercial hosted Polar Signals product launched in beta back in February, and there it shall remain until next year. Branczyk said that the company will eventually offer additional enterprise-grade features, such as automatic recommendations to address infrastructure configuration and code.

Polar Signals’ early user base includes companies that “run foundational pieces of the internet,” such as content delivery networks (CDNs), SaaS companies, database platforms, and even ecommerce companies such as Zalando.

“Our early users find it most useful for saving cost on a cloud bill, and it has shown that most companies are leaving an easy 20% savings on the table because they don’t have insight into what to optimize,” Branczyk said.

Large companies with high cloud bills are Polar Signals’ target

To some, saving on cloud costs might sound like something that younger, cash-strapped startups would be most interested in. But as Branczyk points out, it’s actually larger companies that stand to benefit the most.

“Typically the larger the company, the larger their cloud bill is, so those companies have more to gain, therefore medium to large enterprises are our perfect customers,” Branczyk explained. “Small companies with cloud bills tend to be early-stage startups that don’t really care about their cloud bill efficiency — yet — so those are less likely to be our target customer base.”

Of course, continuous profiling isn’t purely about saving cloud costs — customers expect software to be fluid and fast, so it’s ultimately about improving the overall user experience too.

With $4 million in the bank from big-name backers like GV and Lightspeed, Polar Signals is now well-financed to double down on Parca development and prepare the core commercial hosted product for launch in early 2022.

“Our mission is to not just observe but to truly understand production systems,” Branczyk added. “We feel continuous profiling shines a light on aspects that have been lacking in the observability space, and we have many ideas to further extend our understanding of running systems beyond continuous profiling. We want observability to become understandability.”