GhostPlane is a high-performance eBPF/XDP load balancer and cloud gateway. It leverages XDP for zero-copy L4 steering and Go for advanced L7 routing. Features include a real-time Sankey packet visualizer, Maglev consistent hashing, and an AI-powered Copilot for natural language configuration. Engineered for sub-microsecond latency and 10M+ PPS.
Hardware-level performance using the kernel's fastest path. GhostPlane intercepts packets at the NIC driver level, bypassing the heavy Linux networking stack for sub-microsecond latency.
- Maglev Hashing: Consistent hashing for stable backend distribution.
- Connection Tracking: Stateful flow affinity via LRU maps.
- Dynamic Port Forwarding: Reconfigure listeners in microseconds without reloads.
Smart routing at the application layer with a Go-based control plane.
- Host & Path Based Routing: Fine-grained traffic steering.
- Migration Utility: Automated import tool that converts Nginx/Apache configurations into GhostPlane rules.
Configuring infrastructure shouldn't require a manual. Our integrated LLM assistant helps you:
- Natural Language Config: "Route all /api/v1 traffic to the auth-cluster and enforce a 500req/s limit."
- Root Cause Analysis: Correlates XDP drop stats with backend health to tell you why traffic is failing.
Wait for your logs, or watch your packets. GhostPlane features a high-fidelity Sankey-style visualizer showing real-time traffic distribution from ingress to backend.
graph TD
A[Internet] -->|XDP_DROP/TX| B(XDP Data Plane)
B -->|Bypass Stack| C[Backend Pool A]
B -->|XDP_PASS| D[Linux Network Stack]
D --> E[GhostPlane L7 Proxy]
E --> F[Backend Pool B]
G[Go Control Plane] -->|Map Update| B
H[AI Copilot] -->|Config Gen| G
I[Next.js UI] -->|Dashboard| G
- Linux Kernel 5.15+ (for XDP and BTF support)
- Clang/LLVM 11+
- Go 1.21+
# Build the binary and eBPF objects
go generate ./...
go build -o ghostplane cmd/nlb/main.go
# Run (requires root for eBPF map access)
sudo ./ghostplane --config config.yaml
<!-- slide -->

<!-- slide -->

I am actively looking for contributors to help build the future of cloud-native networking! Whether you are a kernel wizard (eBPF/C), a Go expert, or a Frontend specialist (Next.js), there is a place for you in the cockpit.
Looking for help with:
- Implementing the Maglev hashing C logic.
- Refining the AI RCA engine.
- Kubernetes Ingress Controller integration.
Feel free to open an issue, submit a PR, or reach out!
Maintained by: arunsoman
MIT & Dual-GPL (for eBPF components).
