How LoongCollector Transforms iLogtail into a Next‑Gen Cloud‑Native Observability Agent

Historical Overview

Early computers in the 1960s output status and error logs via printers, laying the foundation for modern observability. Over decades, logging and monitoring evolved from hardware checks to system, network, application, cloud, and container monitoring.

In the past ten years, open‑source has flourished in China, producing projects such as iLogtail, RocketMQ, Nacos, and TiDB.

iLogtail Origin and Milestones

iLogtail was created in 2013 at Alibaba to provide a lightweight, high‑performance, reliable observability data collector. After rigorous testing on Alibaba Cloud and the Alibaba Group, it was fully open‑sourced in 2022 and can run on physical machines, VMs, and Kubernetes, collecting logs, metrics, traces, and profiles.

Since open‑sourcing, iLogtail has released 22 versions, evolving from v1.x to v2.x, adding features such as a unified C++/Golang pipeline event model, native and extensible plugins, SPL programmable syntax, advanced Kubernetes auto‑tagging, OpenTelemetry compatibility, and a ConfigServer for large‑scale configuration management.

Community Growth

The community now has over 1.6K stars, 86 contributors, 7 committers, and 6 PMC members, with 274 issues resolved and 909 pull requests merged. Notable contributors include Liu Haoyang, Zhu Shunjia, Guo Gangping (ByteDance), Sun Yu (Tongcheng Travel), Qiu Fengshuo, and many others who have added plugins, flusher integrations, and performance optimizations.

Brand Evolution to LoongCollector

To celebrate the two‑year anniversary, iLogtail is rebranded as LoongCollector, symbolizing the Chinese dragon’s keen insight, flexibility, and strength. LoongCollector is positioned as a high‑performance, reliable, programmable collector for next‑generation observability pipelines.

Key Features

Uncompromised Performance and Reliability

Designed for large‑scale distributed systems, LoongCollector offers efficient data collection, low latency, and stability under extreme loads.

Unlimited Telemetry

It provides an All‑in‑One agent that captures logs, metrics, traces, events, and profiles, with future plans to enhance Prometheus scraping and integrate eBPF for native metric collection.

Programmable Pipeline

Supports SPL and multi‑language plugins, enabling flexible data processing, custom operators, and composable pipelines.

Unburdened Config Management

Introduces a standardized Agent control protocol and ConfigServer service to centrally manage configurations, monitor agent health, and support distributed storage adapters.

Deployment Modes

Agent mode: runs on hosts or Kubernetes nodes for local, real‑time processing.

Cluster mode: deployed as a service to aggregate data from agents and external protocols.

Stream consumer mode: works with message queues for buffered, scalable stream processing.

Industry Comparison

Compared with Fluent Bit, OpenTelemetry Collector, and Vector, LoongCollector emphasizes comprehensive OneAgent capabilities, high performance, flexible pipelines, and large‑scale configuration control.

Roadmap

Enhance framework for higher performance and reliability (unified send framework, full C++/Golang pipeline, fan‑out, memory control).

Expand All‑in‑One collection (stdout C++ logs, multi‑path file plugins, eBPF network/HTTP/profiling, native Node metrics, Prometheus exporter).

Boost programmable computing (tag handling, SPL aggregation, log‑to‑metric plugins).

Improve usability via ConfigServer upgrades, richer metrics, and Prometheus support.

Broaden ecosystem integrations with more message queues and storage systems.

Conclusion

LoongCollector aims to become a top‑tier cloud‑native observability pipeline, inviting developers and enterprises to join its open‑source community and contribute to its continuous evolution.