LATS

lats is a experimental benchmarking tool NATS JetStream performance. It allows you to measure throughput and latency for different publishing scenarios. There are two publishing modes:

• Batch mode is the same as nats bench js pub async
• Pipelined mode publishes a given number of in-flight messages, and when an acknowledgment is is received, a new message is published immediately.

Usage

lats [flags]

Flags

Flag	Description	Default
-url	NATS server URL	nats://127.0.0.1:4222
-stream	JetStream stream name	BENCH
-subject	Subject to publish messages to	bench.test
-size	Message size in bytes	128
-clients	Number of concurrent publisher threads	1
-batch	Max batch size per publisher (for batch mode)	500
-inflight	Max outstanding async publishes per publisher (for pipelined mode). If 0, batch publishing is used.	0
-duration	Benchmark duration	10s
-replicas	Number of stream replicas	3
-sysurl	URL for system account connection to find the stream leader (e.g. nats://user:pass@host:port)	""

Example

Run a benchmark with 1 publisher, a batch size of 1000, for 30 seconds:

lats -clients 1 -batch 1000 -duration 30s

Run a benchmark in pipelined mode with 16 in-flight messages:

lats -clients 1 -inflight 16 -duration 30s

Sample Output

The tool outputs the benchmark parameters were used and the results in JSON format.

{
  "params": {
    "publishers": 1,
    "inflight": 0,
    "batch": 1,
    "message_size_bytes": 128,
    "duration_sec": 60
  },
  "results": {
    "total_messages": 614323,
    "throughput_msg_per_sec": 10238,
    "latency_us": {
      "min": 58,
      "P50": 95,
      "P90": 109,
      "P99": 130,
      "P99.9": 213,
      "max": 7654
    }
  }
}