Continuous profiling is now GA

  • 5 min read

Dash0 Announces General Availability of Continuous Profiling for CPU, Memory and Any Profile in OpenTelemetry Format

Evgeni WachnowezkiEvgeni Wachnowezki
Frederic BranczykFrederic Branczyk
Dash0 announces general availability of continuous profiling for CPU, memory and any profile in OpenTelemetry format, illustrated with a flame graph and CPU profiling data.

Dash0's always-on, zero-instrumentation profiler shows the line of code behind every CPU spike and memory leak, takes any profile in the OpenTelemetry format, and shows it next to the traces, logs and metrics from the same pod.

SALT LAKE CITY, KubeCon + CloudNativeCon North America – Dash0, the OpenTelemetry-native agentic observability platform, today announced the general availability of Profiling in Dash0, a continuous profiler that measures CPU and memory for every process in a Kubernetes cluster or on any other Linux host, VM or bare metal, with no code changes, and accepts any profile sent in the OpenTelemetry profiling format. The launch comes 12 weeks after Dash0 acquired Polar Signals, the Berlin-based company behind the open-source Parca project, and makes code-level profiles a native signal alongside traces, logs and metrics. Polar Signals counts Notion, Vercel and Modal among its customers.

Dash0 Profiling runs on every node all the time, with overhead that Polar Signals measured at under 1% even under heavy CPU load, so the profile from the minute something went wrong is already there.

"At last, continuous profiling becomes a first-class signal in observability! When we announced Polar Signals in August, we said profiling would be an integral part of Dash0 capabilities. Twelve weeks later, it is. Every team running on Kubernetes can now see which line of code is behind a CPU spike or a memory leak, right next to the trace and the log that told them something was wrong. And because Agent0 and coding agents see the same profiles, the fix can start the moment the problem shows up," said Mirko Novakovic, CEO and Co-Founder, Dash0.

"We spent years making continuous profiling efficient enough to run on every node, all the time. What was missing was the rest of the picture, just as I predicted in a KubeCon Keynote 2019 in Barcelona. In Dash0, a profile sits next to the span, the log line and the metric from the same pod and the same minute. That is when profiling stops being a specialist's tool and becomes something every engineer uses during an incident - finally fulfilling the vision laid out on the keynote stage in 2019," said Frederic Branczyk, VP of Data Engineering, founder of Polar Signals.

Dash0 Profiling gives platform and application teams:

  • Always-on, zero instrumentation: An eBPF agent on each node profiles every process in every language, from Go, Rust and C/C++ to Java, Python, Node.js, Ruby, PHP and .NET. Profiles captured before and during an incident are there for the postmortem.
  • CPU, memory and any OpenTelemetry profile: CPU shows where compute goes, down to the function and line. Memory shows which call paths allocate and which allocations stay live. Any other profile in the OpenTelemetry profiling format, from a runtime profiler or an SDK, lands in the same view.
  • From trace to profile in one click: Dash0 links eBPF profiles to the traces they belong to. From a slow span, engineers open the profile of exactly that request and see which functions and lines spent the time.
  • Built for agents: Agent0 and coding agents such as Claude Code read profiles through Dash0's MCP server, so suggested fixes rest on measured cost rather than guesswork.
  • Open by design: Polar Signals engineers contribute to the OpenTelemetry Profiling Working Group, whose profiles signal entered Alpha in March 2026, and to the open-source OpenTelemetry eBPF profiler. Parca remains open source.

Profiling is available now to all Dash0 customers for $0.60 per 1M samples. New users can try it in Dash0's 14-day free trial at dash0.com/sign-up. KubeCon attendees can see live demos at booth 134, Salt Palace Convention Center, November 10–12.

About Dash0

Dash0 closes the loop from code to production in the AI era. Built natively on OpenTelemetry, Dash0 tracks what AI builds, observes everything running in production, and fixes problems autonomously, all powered by a single, open data foundation. Dash0 helps engineering teams ship more and break less.

Founded in Germany in 2023 and headquartered in New York, Dash0 serves 850+ customers, including Zalando and The Telegraph. In August 2026, Dash0 acquired Polar Signals, the Berlin-based continuous profiling company behind the open-source Parca project. The company raised $110M in Series B funding at a $1B valuation, with participation from Balderton Capital, Accel, Cherry Ventures, DTCP, DIG Ventures, July.fund and T.Capital. Learn more at dash0.com.

Media contact: Becki Howland, becki.howland@dash0.com