LoongCollector - High-Performance Observability Data Collector

Fast, Lightweight, and Battle-Tested Observability Agent

Quick Start • Documentation • Performance Benchmarks • Contributing

---

Why LoongCollector?

LoongCollector is a high-performance, lightweight observability data collector designed for modern cloud-native environments. Born from Alibaba's 15-year journey handling exponential traffic growth and powering tens of millions of deployments, LoongCollector delivers 10x higher throughput with 80% lower resource usage compared to open-source alternatives.

🚀 Core Advantages

• ⚡ High Performance

  • 10x higher throughput with 80% less resource usage compared to competitors
  • Linear scaling with predictable performance growth

• 🛡️ Production-Ready

  • Battle-tested in Alibaba's production environment for 15+ years
  • Powers tens of millions of deployments, collecting hundreds of petabytes daily

• 🔧 All-in-One Collection

  • Unified agent for logs, metrics, traces, events, and profiles
  • Native Kubernetes support
  • eBPF-powered network monitoring and security event collection

• 🎯 Pluggable Architecture

  • 100+ built-in plugins with multi-language development (C++, Go)
  • Powerful SPL engine for flexible data processing

• ⚙️ Advanced Management

  • Remote configuration management via SLS console, SDK, K8s Operator
  • Self-monitoring, flow control, resource control, alarms, and statistics collection.

📊 Performance Benchmarks

Maximum Throughput Comparison

Log Type	LoongCollector	FluentBit	Vector	Filebeat
Single Line	546 MB/s	36 MB/s	38 MB/s	9 MB/s
Multi-line	238 MB/s	24 MB/s	22 MB/s	6 MB/s
Regex Parsing	68 MB/s	19 MB/s	12 MB/s	Not Supported

📈 Breaking Point Analysis: While competitors hit CPU saturation at ~40 MB/s, LoongCollector maintains linear scaling up to 546 MB/s on a single processing thread.

Resource Efficiency at 10 MB/s Processing Load

Scenario	LoongCollector	FluentBit	Vector	Filebeat
Simple Line (512B)	3.40% CPU 29.01 MB RAM	12.29% CPU (+261%) 46.84 MB RAM (+61%)	35.80% CPU (+952%) 83.24 MB RAM (+186%)	Performance Insufficient
Multi-line (512B)	5.82% CPU 29.39 MB RAM	28.35% CPU (+387%) 46.39 MB RAM (+57%)	55.99% CPU (+862%) 85.17 MB RAM (+189%)	Performance Insufficient
Regex (512B)	14.20% CPU 34.02 MB RAM	37.32% CPU (+162%) 46.44 MB RAM (+36%)	43.90% CPU (+209%) 90.51 MB RAM (+166%)	Not Supported

🏗️ Performance And Reliability Architecture Highlights

1. Memory Arena: Zero-Copy Design

• Shared memory pool (SourceBuffer) stores all string data once per event group
• String_view references point to original data segments instead of copying

2. Lock-Free Event Pool

• Thread-aware allocation strategies eliminate lock contention
• Same-thread pools for direct reuse, double-buffer pools for cross-thread scenarios

3. Zero-Copy Serialization: Direct Network Output

• Bypasses intermediate Protobuf objects, serializes directly to network format

4. Multi-Tenant Pipeline Isolation

• High-low watermark feedback queues prevent pipeline interference
• Independent resource allocation per pipeline with automatic back-pressure control
• Ensures one pipeline failure doesn't affect others

5. Fair Resource Allocation

• Priority-aware round-robin scheduling ensures fairness while respecting business priorities
• Higher priority pipelines always processed first, fair distribution within same priority level
• Automatic resource yielding when constraints occur

6. Self-Healing Network Resilience

• Adaptive concurrency limiting per destination using AIMD (Additive Increase, Multiplicative Decrease)
• Fast failure detection and gradual recovery to prevent network jitter
• Zero data loss guarantee with intelligent back-pressure control

🏭 Production Validation: Battle-Tested at Scale

LoongCollector has been battle-tested in some of the world's most demanding production environments:

• Alibaba Group: Powers the entire Alibaba ecosystem including Double 11 shopping festival
• Alibaba Cloud: Serves tens of thousands of enterprise customers
• Ant Group: Handles financial transaction observability at massive scale
• Daily Data Volume: Hundreds of petabytes of observability data
• Deployment Scale: Tens of millions of active deployments

🚀 Quick Start

Prerequisites

• Docker (for building from source)
• Go 1.19+ (for building from source)

Build and Run

# Clone the repository
git clone https://github.com/alibaba/loongcollector.git
cd loongcollector

# Build LoongCollector
make all
cd output

# Start LoongCollector
nohup ./loongcollector > stdout.log 2> stderr.log &

LoongCollector is now running.

Docker Quick Start

# Build the Docker image alibaba/loongcollector:0.0.1
make dist
make docker

# Run with default configuration
docker run -d --name loongcollector \
  -v /:/logtail_host:ro \
  -v /var/run:/var/run \
  alibaba/loongcollector:0.0.1

📚 Documentation

• User Manual - Comprehensive documentation
• Installation Guide - Step-by-step setup
• Configuration Reference - Detailed configuration options
• Plugin Directory - Complete plugin documentation
• Developer Guide - Contributing and development

🤝 Contributing

We welcome contributions from the community! Here are some ways you can help:

• Report Bugs - Help us identify and fix issues
• Improve Documentation - Enhance our docs
• Review Code - Review feature proposals and PRs
• Contribute Plugins - Develop new input, processor, or flusher plugins

📞 Contact Us

• GitHub Issues: Report bugs and feature requests
• GitHub Discussions: Join community discussions

Community Channels

• Bilibili: 阿里云SLS
• Zhihu: LoongCollector社区
• WeChat/DingTalk: LoongCollector社区

📄 License

LoongCollector is licensed under the Apache 2.0 License.

---

Built with ❤️ by the Alibaba Cloud Observability Team

Empowering developers to build better observability solutions