libdd-profiling

Core profiling library for collecting, aggregating, and exporting profiling data in pprof format to Datadog.

Overview

libdd-profiling provides the core functionality for continuous profiling, including profile collection, aggregation, compression, and export to Datadog backends using the pprof format.

Features

• Profile Management: Collect and manage profiling data (CPU, memory, allocations, etc.)
• Sample Aggregation: Efficiently aggregate samples with stack traces
• pprof Format: Generate profiles in Google's pprof protobuf format
• Compression: LZ4 compression for efficient data transfer
• Stack Traces: Full stack trace capture with mapping and function information
• Value Types: Support for multiple value types (CPU time, memory, count, etc.)
• Upscaling: Statistical upscaling for sampled data
• HTTP Export: Built-in HTTP exporter with multipart form data support

Modules

• api: Core API types (ValueType, Period, Mapping, Function, etc.)
• collections: String storage and interning for efficient memory use
• exporter: HTTP exporter for sending profiles to Datadog
• internal: Internal profile management and aggregation
• iter: Iteration utilities for profile data
• pprof: pprof protobuf format support

Example Usage

use libdd_profiling::api::{Profile, ValueType};

// Create a profile
let value_types = vec![
    ValueType::new("samples", "count"),
    ValueType::new("cpu", "nanoseconds"),
];

// Add samples with stack traces
// ... collect profiling data ...

// Export to Datadog
// ... use exporter to send profile ...

Profile Format

The library generates profiles in the pprof format, which includes:

• Stack traces with function names and locations
• Sample values (counts, times, sizes)
• Mappings for binaries and shared libraries
• Labels for additional context