Pyroscope 2.0: Accelerating Continuous Profiling with Enhanced Scalability and OTLP Support

Why Continuous Profiling Is Becoming Indispensable

Continuous profiling is rapidly evolving into a fundamental pillar of observability, and for compelling reasons. Unlike other signals, profiling reveals why code performs poorly—not just that it does. Metrics highlight high CPU usage; logs flag slow requests; traces pinpoint the bottleneck service. But only a profile shows exactly which function and line of code is consuming resources.

As distributed systems grow in complexity, this depth of insight becomes critical. The OpenTelemetry project recently marked its Profiles signal as alpha, signaling profiling’s rise to first‑class citizen status in the observability ecosystem. Now, with the release of Pyroscope 2.0, our open source continuous profiling database has been rebuilt from the ground up to make always‑on profiling more cost‑effective at scale—and with native support for the OpenTelemetry Protocol (OTLP) for profiling, teams can start ingesting profiles using the emerging standard today.

The Tangible Benefits of Always‑On Profiling

Before diving into Pyroscope 2.0’s new capabilities, it’s worth exploring why continuous profiling delivers outsized value across engineering teams.

Cut Infrastructure Costs with Data, Not Guesswork

Cloud spending often dominates engineering budgets, with CPU and memory representing a significant portion. Teams routinely over‑provision because they lack granular, production‑level visibility into resource consumption.

Continuous profiling changes this. By seeing exactly which functions drive CPU and memory usage—across every service, in production, over time—engineers can make targeted optimizations rather than blindly scaling hardware. The result: lower infrastructure costs driven by real data, not hunches.

Faster Root Cause Analysis

When an incident strikes, the first question is always why. Metrics and traces narrow the blast radius—you know which service, endpoint, or deployment introduced a regression. But the “last mile” of root cause analysis often consumes hours.

With continuous profiling, that last mile shrinks to minutes. Compare a profile from before and after the regression, diff them, and immediately see which code paths changed. No need to reproduce in staging, add ad‑hoc logging, or guess.

Understand Latency at the Code Level

Distributed tracing shows where wall‑clock time is spent; profiling shows where the CPU actually spends that time. Together, they close the observability gap. For instance, a trace might reveal that your auth service added 200 ms to a request, while a profile shows that 150 ms of that came from an uncached regex compilation.

This insight is especially powerful for tail latency. The p99 spikes that are notoriously hard to reproduce and diagnose are captured automatically by continuous profiling—no need to rely on luck with a debugger.

What’s New in Pyroscope 2.0

Pyroscope 2.0 represents a complete rearchitecture of our open source continuous profiling database, designed to make profiling more scalable and cost‑effective than ever before.

A Rearchitected Foundation for Scale

The original Pyroscope was built on the Cortex foundation (the same base used by Mimir and Loki). While functional, that architecture had limitations when handling profiling data at very large scales. Pyroscope 2.0 introduces a new storage engine optimized for the high‑cardinality, high‑frequency nature of profiling data. This redesign delivers significant improvements in ingestion throughput, query performance, and storage efficiency.

Native OTLP Profiling Support

With OpenTelemetry’s Profiles signal reaching alpha, the industry is rallying around a standard for profiling data. Pyroscope 2.0 embraces this by offering native support for the OpenTelemetry Protocol (OTLP) for profiles. This means you can start sending profiling data using the same instrumentation and collectors you already use for traces and metrics—reducing the need for custom agents and simplifying your observability stack.

Cost‑Effective Operations at Scale

Continuous profiling is only practical if it doesn’t break the bank. Pyroscope 2.0’s rearchitecture includes improvements in data compression, retention management, and query efficiency. As a result, teams can run always‑on profiling across thousands of services without runaway storage or compute costs. The new architecture also supports multi‑tenancy out of the box, making it easy to share a single profiling backend across multiple teams or environments.

The Future of Profiling in Observability

With OpenTelemetry standardizing the Profiles signal and Pyroscope 2.0 delivering a production‑ready, cost‑effective database, continuous profiling is poised to become as routine as metrics, logs, and traces. Teams that adopt it today will gain a decisive advantage in understanding—and optimizing—their systems from the inside out.