deepflowio/deepflow
eBPF Observability - Distributed Tracing and Profiling
What it solves
DeepFlow provides deep observability for complex cloud-native and AI applications. It eliminates the need for developers to manually instrument code, removing the heavy burden of continuous manual monitoring setup for DevOps and SRE teams.
How it works
DeepFlow uses eBPF and Wasm plugins to implement "Zero Code" data collection for metrics, distributed tracing, request logs, and function profiling. It employs a technique called SmartEncoding to pre-encode meta tags, which reduces storage overhead by 10x compared to standard ClickHouse methods and allows for high-dimensional data queries without performance loss.
Who it’s for
It is designed for DevOps and SRE teams managing cloud-native and AI applications, as well as developers who need full-stack observability across application services, AI services, and infrastructure.
Highlights
- Zero Code Instrumentation: Automatically collects metrics and tracing data using eBPF without requiring code changes.
- Full Stack Correlation: Correlates data from the application, runtime, kernel, and infrastructure (gateways, databases, message queues).
- Continuous Profiling: Collects OnCPU/OffCPU/GPU/Memory/Network flame graphs with less than 1% overhead.
- Broad Integration: Acts as a storage backend for Prometheus, OpenTelemetry, SkyWalking, and Pyroscope.
- High Efficiency: Uses SmartEncoding to handle unlimited dimensions and cardinalities with significantly reduced storage costs.