Building Large-Scale Service Topology: Architecture, Challenges, and Lessons

Netflix engineers needed to build a real-time service dependency map to understand the blast radius of failures within their distributed architecture and quickly resolve issues.

To achieve this, they adopted a multi-source approach combining eBPF network flows, IPC metrics, and distributed tracing data. This data is collected and processed through a streaming architecture.

The key technical challenges included:

• Building a distributed aggregation pipeline to process large-scale data streams
• Implementing Time-travel queries to look up past point-in-time states
• Designing a structure to visualize and explore the complex relationships between numerous services

This project goes beyond simple monitoring, providing a foundation for engineers to understand and respond to the connectivity of the entire system in real time.