Course contentsShow
← System Design
Lesson 1242 of 1,91931. Observability: TracingFree lesson

Zipkin Architecture and Design

Explore Zipkin's simpler architecture, storage backends, and how it differs from Jaeger in deployment complexity.

Zipkin Architecture and Design

What you'll learn: How Zipkin's simpler architecture and flexible storage approach differ from Jaeger's deployment model.

A Simpler Tracing System

Zipkin was one of the first open-source distributed tracing systems (inspired by Google's Dapper). Unlike Jaeger's multi-component architecture with separate collectors, agents, and query services, Zipkin takes a minimalist approach: at its core, you can run Zipkin as a single executable that handles trace ingestion, storage, and querying all in one process.

Think of Zipkin like an all-in-one printer (print, scan, fax in one box), while Jaeger resembles a specialized printing department with separate machines for each function.

Core Components

Zipkin's architecture consists of:

  1. Collector — receives span data from instrumented applications via HTTP, Kafka, or RabbitMQ
  2. Storage — persists traces (in-memory, MySQL, Cassandra, or Elasticsearch)
  3. Query Service — API for retrieving traces
  4. UI — web interface for visualization

The key difference: these can all run in a single process, making Zipkin incredibly easy to deploy for smaller systems. You can literally start Zipkin with one JAR file.

Storage Flexibility

Zipkin supports multiple storage backends out-of-the-box:

  • In-memory (ephemeral, good for testing)
  • MySQL (simple, familiar relational DB)
  • Cassandra (scalable, distributed)
  • Elasticsearch (full-text search, popular choice)

This pluggable storage makes Zipkin adaptable to different operational preferences.

Deployment Complexity Comparison

Zipkin wins on simplicity: single binary deployment works for many use cases. Jaeger wins on scalability: its separated components (agent, collector, query) provide better horizontal scaling and fault isolation for large-scale systems. Zipkin's monolithic option means one failure point affects everything; Jaeger's architecture isolates failures better.

Key Takeaway: Zipkin offers a simpler, monolithic architecture with flexible storage backends, making it easier to deploy for small-to-medium systems, while Jaeger's component-based design provides better scalability and resilience for large-scale distributed environments.