lavinmqperf
lavinmqperf is a built-in benchmarking tool for measuring LavinMQ performance.
Usage
lavinmqperf [protocol] [scenario] [options]
protocol—amqp(default) ormqttscenario— one of the scenarios listed belowoptions— global options plus scenario-specific options
If no protocol is given, amqp is assumed. Run lavinmqperf with no arguments to see the banner with available scenarios.
Examples:
lavinmqperf amqp throughput --publishers=4 --consumers=4 --size=1024 --time=60
lavinmqperf mqtt throughput --qos=1 --time=30 --uri=mqtt://localhost:1883
lavinmqperf amqp queue-count --queues=10000 --durable
Global Options
These apply to every scenario.
| Flag | Description |
|---|---|
--uri=URI |
AMQP or MQTT URI. Default: amqp://guest:guest@localhost for AMQP scenarios, mqtt://localhost:1883 for MQTT scenarios. |
-h, --help |
Show help |
-v, --version |
Show version |
--build-info |
Show build information |
AMQP Scenarios
throughput — publish and consume throughput
Spawns publishers and consumers and measures sustained publish and consume rates. Optionally measures end-to-end latency.
| Flag | Description |
|---|---|
-x, --publishers=N |
Number of publishers (default 1) |
-y, --consumers=N |
Number of consumers (default 1) |
-s, --size=BYTES |
Message body size (default 16) |
-V, --verify |
Verify the message body on consume |
-a, --ack=N |
Ack every N consumed messages (default 0 = no-ack mode) |
-c, --confirm=N |
Confirm publishes; max N unconfirmed in flight |
-t, --transaction=N |
Publish in transactions, committing every N messages |
-T, --transaction=N |
Ack in transactions, committing every N messages |
-u, --queue=NAME |
Queue name (default perf-test). Also used as the routing key. |
-k, --routing-key=KEY |
Routing key (default queue name) |
-e, --exchange=NAME |
Exchange to publish to (default "", i.e. the default exchange) |
-r, --rate=N |
Max publish rate per publisher in msgs/s (default 0 = unlimited) |
-R, --consumer-rate=N |
Max consume rate per consumer in msgs/s (default 0 = unlimited) |
-p, --persistent |
Mark messages as persistent (delivery_mode=2) |
-P, --prefetch=N |
Consumer prefetch count |
-g, --poll |
Use basic.get polling instead of basic.consume |
-j, --json |
Print final summary as JSON |
-z, --time=SECONDS |
Stop after this many seconds |
-q, --quiet |
Only print the final summary |
-C, --pmessages=N |
Stop publishers after publishing N messages total |
-D, --cmessages=N |
Stop consumers after consuming N messages total |
--queue-args=JSON |
Queue declaration arguments as a JSON object |
--consumer-args=JSON |
Consumer arguments as a JSON object |
--properties=JSON |
Message properties as a JSON object |
--random-bodies |
Each message body is randomized |
--queue-pattern=PATTERN |
Use multiple queues. % is replaced with a number (e.g. queue-%). |
--queue-pattern-from=N |
Start index for --queue-pattern (default 1) |
--queue-pattern-to=N |
End index for --queue-pattern (default 1) |
-l, --measure-latency |
Measure end-to-end latency. Requires --size>=8. The first 8 bytes of each message carry a timestamp. |
queue-count — performance with many queues
Declares many queues and reports declaration time.
| Flag | Description |
|---|---|
-q, --queues=N |
Number of queues (default 100) |
-D, --durable |
Declare the queues as durable |
--arguments=JSON |
Queue arguments as a JSON object |
-n, --no-wait |
Use nowait so the client does not wait for declaration confirmation |
connection-count — performance with many connections
Opens many connections (and optionally channels and consumers per connection).
| Flag | Description |
|---|---|
-x, --count=N |
Number of connections (default 100) |
-c, --channels=N |
Channels per connection (default 1) |
-C, --consumers=N |
Consumers per channel (default 0) |
-u, --queue=NAME |
Queue name to consume from |
-l, --localhost |
Connect to a random 127.0.0.0/16 address (useful to spread source ports) |
-k, --keepalive=IDLE:COUNT:INTERVAL |
TCP keepalive parameters |
bind-churn — binding create/delete rate
Measures queue.bind throughput against amq.direct. Uses Benchmark.ips. No scenario-specific flags; only the global options apply.
queue-churn — queue create/delete rate
Measures transient and durable queue create/delete throughput. Uses Benchmark.ips. No scenario-specific flags.
connection-churn — connection open/close rate
Measures connection open/close throughput. Uses Benchmark.ips. No scenario-specific flags.
channel-churn — channel open/close rate
Measures channel open/close throughput on a single connection. Uses Benchmark.ips. No scenario-specific flags.
consumer-churn — consumer subscribe/cancel rate
Measures basic.consume / basic.cancel throughput. Uses Benchmark.ips. No scenario-specific flags.
MQTT Scenarios
throughput — MQTT publish and subscribe throughput
| Flag | Description |
|---|---|
-u, --uri=URI |
MQTT broker URI (default mqtt://localhost:1883). Overrides the global --uri. |
-x, --publishers=N |
Number of publishers (default 1) |
-y, --consumers=N |
Number of consumers (default 1) |
-s, --size=BYTES |
Message body size (default 16) |
-V, --verify |
Verify the message body on consume |
-q, --qos=LEVEL |
QoS level: 0 or 1 |
-t, --topic=NAME |
Topic name (default perf-test) |
-r, --rate=N |
Max publish rate per publisher in msgs/s (default 0 = unlimited) |
-R, --consumer-rate=N |
Max consume rate per consumer in msgs/s (default 0 = unlimited) |
-j, --json |
Print final summary as JSON |
-z, --time=SECONDS |
Stop after this many seconds |
-q, --quiet |
Only print the final summary |
-C, --pmessages=N |
Stop publishers after publishing N messages total |
-D, --cmessages=N |
Stop consumers after consuming N messages total |
--random-bodies |
Each message body is randomized |
--retain |
Set the retain flag on published messages |
--clean-session |
Connect with clean_session=true |
Output
By default, the throughput scenarios print one line per second with current publish/consume rates and (if --measure-latency is set) a percentile snapshot. On stop, they print a summary with average rates and aggregate latency percentiles.
Use -j / --json for machine-readable output:
{
"elapsed_seconds": 30.0,
"avg_pub_rate": 125000,
"avg_consume_rate": 124998,
"latency_min_ms": 0.412,
"latency_median_ms": 0.987,
"latency_p75_ms": 1.215,
"latency_p95_ms": 2.043,
"latency_p99_ms": 4.871,
"latency_count": 3749940,
"latency_sample_size": 131072
}
The latency reservoir holds up to 131,072 samples; latency_count is the total number of measured deliveries.
Stopping a Run
Send SIGINT (Ctrl+C) once for a graceful stop with a final summary. Send it twice to abort immediately.
Ready to take the next steps?
Managed LavinMQ instance via CloudAMQP
LavinMQ has been built with performance and ease of use in mind - we've benchmarked a throughput of about 1,000,000 messages/sec . You can try LavinMQ without any installation hassle by creating a free instance on CloudAMQP. Signing up is a breeze.
Get started with CloudAMQP ->
Help and feedback
We welcome your feedback and are eager to address any questions you may have about this piece or using LavinMQ. Join our Slack channel to connect with us directly. You can also find LavinMQ on GitHub.