Aeron Objectives and Components

Aeron is built around three non-negotiable performance goals, and it ships as a layered stack of three components. Understand both, and every tuning decision later in this guide has a place to hang.

The three objectives

Aeron’s design is driven by three goals that cannot be traded away:

1. Ultra-low latency — Minimize the time from message send to receive, targeting single-digit microseconds.
2. High throughput — Sustain millions of messages per second per stream.
3. Predictable latency (p99 and beyond) — Tail latency matters more than the average. A system that is fast on average but spikes at p99 is unreliable for trading workloads.

The first two goals are about speed and volume. The third is about consistency — and it is the one that defines Aeron.

Key Insight: the tail is what kills you

In financial systems, p99/p999 latency is what kills you — not the average. Aeron is purpose-built to tame tail latency through mechanical sympathy, zero-copy design, and lock-free data structures.

Keep this lens on every knob in the Performance Tuning section: a setting that lowers p50 but widens the gap to p99 is usually the wrong trade for a trading workload. Reducing baseline cost helps p50; eliminating variance — cache misses, queue depth, context switches — is what protects p99.

The three components of Aeron OSS

The open-source edition has three interconnected components arranged in a layered architecture:

┌─────────────────┐
│  Aeron Cluster   │  ← Consensus + state machine replication (top layer)
├─────────────────┤
│  Aeron Archive   │  ← Persistent recording & replay (middle layer)
├─────────────────┤
│  Aeron Transport │  ← Reliable UDP messaging (foundation)
└─────────────────┘

• Aeron Transport sits at the base — all communication flows through it.
• Aeron Archive builds on Transport to add stream recording and replay.
• Aeron Cluster sits on top, using both Transport and Archive to implement Raft consensus.

Layer	What it does	Builds on
Aeron Cluster	Consensus and state machine replication via Raft	Transport + Archive
Aeron Archive	Persistent recording and replay of streams	Transport
Aeron Transport	Reliable UDP messaging	— (standalone)

Key Insight: dependency flows upward

The arrows in the diagram show dependency flow: Cluster depends on Archive, Archive depends on Transport. You can use Transport standalone, but Cluster requires all three.

Where to go deeper

This page is the map, not the territory. For the deep internals — how the transport protocol, log buffers, archive recording, and the Raft election state machine actually work — read The Aeron Files. This guide stays thin on internals on purpose and focuses on the operational side: tuning, measuring, monitoring, and recovering.