DeepFlow provides zero-code, full-stack observability for cloud-native and AI applications using eBPF and Wasm. It targets DevOps/SRE teams and developers, offering automated data collection for metrics, distributed tracing, and profiling, correlating data from code to infrastructure without manual instrumentation.

How It Works

DeepFlow leverages eBPF for kernel-level data collection, enabling zero-code instrumentation of applications and infrastructure. Its SmartEncoding technology pre-encodes metadata, reducing storage by 10x compared to traditional methods like ClickHouse, while supporting high-cardinality tags for deep analysis. Wasm plugins extend protocol support for custom or private protocols.

Quick Start & Requirements

• Installation: Deployment documentation available for Community and Enterprise editions. A demo environment is provided.
• Prerequisites: eBPF technology, Kubernetes (for Server component).
• Resources: Agent runs on each node; Server runs in a Kubernetes cluster.

Highlighted Details

• Zero-code distributed tracing and continuous profiling (OnCPU/OffCPU/GPU/Memory/Network) with <1% overhead.
• Universal service mapping for any language and infrastructure, including AI services.
• 10x storage reduction via SmartEncoding, offering BigTable-like query performance.
• Seamless integration with Prometheus, OpenTelemetry, SkyWalking, and Pyroscope.

Maintenance & Community

• Active development with a public roadmap.
• Community support via Discord and WeChat.
• Paper accepted by ACM SIGCOMM 2023.

Licensing & Compatibility

• License: Apache 2.0.
• Compatibility: Compatible with commercial and closed-source applications.

Limitations & Caveats

The Enterprise demo is currently available in Chinese only.