Aeron is a high-performance messaging system designed for low-latency, high-throughput, and reliable delivery in environments like financial trading, real-time analytics, and distributed systems.

It was originally developed by Real Logic and is open-source (Apache 2.0 license). Aeron is used in situations where traditional messaging systems (e.g., Kafka, RabbitMQ, ZeroMQ) are either too heavyweight or introduce too much latency.

Core Features

• Transport Layer

  • Aeron provides both UDP unicast and multicast transport, as well as IPC (inter-process communication) over shared memory.
  • Can be used between threads, processes, and machines with a unified abstraction.

• High-Performance Design

  • Designed for consistent sub-100 microsecond latency, even at millions of messages per second.
  • Uses lock-free, non-blocking structures (like ring buffers) and memory-mapped files.

• Reliable Messaging with Media Drivers

  • The Media Driver handles all the socket-level operations.
  • Clients talk to the media driver via shared memory, reducing syscall overhead.

• Back Pressure Handling

  • Built-in mechanisms to apply flow control and back pressure to producers if consumers can’t keep up.

• Loss-Tolerant or Reliable Modes

  • UDP itself is unreliable, but Aeron offers its own reliability features—either loss-tolerant (e.g., for market data) or guaranteed delivery with retransmits and acknowledgments (via Aeron’s Reliable Transport or higher-level abstraction called Aeron Archive).

• Cluster Mode (Aeron Cluster)

  • Aeron Cluster provides a Raft-based consensus and replicated log, ideal for building leader-follower systems or replicated state machines.

Key Components

• Aeron Client – used by applications to publish or subscribe to message streams.
• Media Driver – separate process (or embedded) that handles actual I/O.
• Aeron Archive – optional component for persistent messaging and replay (like Kafka).
• Aeron Cluster – a Raft-based distributed consensus system built on Aeron messaging.

Typical Use Cases

• Market data distribution
• Order management systems
• High-performance service-to-service messaging
• Replicated logs and state machines
• Real-time analytics

Language Support

• Native Java implementation (most complete)
• C++ and Rust clients available
• Golang and Python have unofficial/experimental support

Comparison to Other Tools