MQTT Bench Guide

This document focuses on robust-bench mqtt usage and benchmark practices.

1. Subcommands

robust-bench mqtt conn ...
robust-bench mqtt pub ...
robust-bench mqtt sub ...

2. conn: Connection Benchmark

Purpose

Measure connection establishment capability and stable connection scale.

--mode semantics:

• create: stop once connection creation completes
• hold: keep connected for --hold-secs
• --concurrency: number of in-flight connection creation tasks

Example

robust-bench mqtt conn \
  --host 127.0.0.1 \
  --port 1883 \
  --count 50000 \
  --concurrency 1000 \
  --mode hold \
  --hold-secs 90

3. pub: Publish Benchmark

Purpose

Measure publish throughput, success rate, and publish latency.

Key Options

• --topic: topic template, supports %i
• --payload-size: payload size in bytes
• --message-interval-ms: per-client publish interval
• --qos: QoS level

Example

robust-bench mqtt pub \
  --host 127.0.0.1 \
  --port 1883 \
  --count 2000 \
  --topic load/%i \
  --payload-size 512 \
  --message-interval-ms 5 \
  --qos 0 \
  --duration-secs 120

4. sub: Subscribe Benchmark

Purpose

Measure receive throughput, stability, and subscriber-side timeout behavior.

Example

robust-bench mqtt sub \
  --host 127.0.0.1 \
  --port 1883 \
  --count 5000 \
  --topic "load/#" \
  --qos 1 \
  --duration-secs 120

5. How to Read the Result

• avg_ops_per_sec: average throughput
• peak_ops_per_sec: peak throughput
• success_rate(%): reliability indicator
• latency_p95/p99: tail latency, critical for performance analysis
• Error Distribution: first place to locate failure reasons

6. Recommended Workflow

1. Run conn first to determine connection capacity.
2. Run pub + sub for data-plane performance.
3. Repeat each scenario three times and use median as baseline.