nats-bench

NATS is fast and lightweight and places a priority on performance. NATS provides tools for measuring performance. In this tutorial you learn how to benchmark and tune NATS on your systems and environment.

Prerequisites

Start the NATS server with monitoring enabled

  1. % nats-server -m 8222

Verify that the NATS server starts successfully, as well as the HTTP monitor:

  1. [18541] 2016/10/31 13:26:32.037819 [INF] Starting nats-server version 0.9.4
  2. [18541] 2016/10/31 13:26:32.037912 [INF] Starting http monitor on 0.0.0.0:8222
  3. [18541] 2016/10/31 13:26:32.037997 [INF] Listening for client connections on 0.0.0.0:4222
  4. [18541] 2016/10/31 13:26:32.038020 [INF] Server is ready

Installing and running the benchmark utility

The NATS benchmark can be installed and run via Go. Ensure your golang environment is setup.

There are two approaches; you can either install the nats-bench utility in the directory specified in your GOBIN environment variable:

  1. % go install $GOPATH/src/github.com/nats-io/nats.go/examples/nats-bench/main.go

… or you can simply run it via go run:

  1. % go run $GOPATH/src/github.com/nats-io/nats.go/examples/nats-bench/main.go

On windows use the % environment variable syntax, replacing $GOPATH with %GOPATH%.

For the purpose of this tutorial, we’ll assume that you chose the first option, and that you’ve added the GOBIN environment variable to your PATH.

The nats-bench utility is straightforward to use. The options are as follows:

  1. % nats-bench -h
  2. Usage: nats-bench [-s server (nats://localhost:4222)] [--tls] [-np NUM_PUBLISHERS] [-ns NUM_SUBSCRIBERS] [-n NUM_MSGS] [-ms MESSAGE_SIZE] [-csv csvfile] <subject>

The options are self-explanatory. Each publisher or subscriber runs in its own go routine with its own NATS connection.

Run a publisher throughput test

Let’s run a test to see how fast a single publisher can publish one million 16 byte messages to the NATS server.

  1. % nats-bench -np 1 -n 100000 -ms 16 foo

The output tells you the number of messages and the number of payload bytes that the client was able to publish per second:

  1. Starting benchmark [msgs=100000, msgsize=16, pubs=1, subs=0]
  2. Pub stats: 7,055,644 msgs/sec ~ 107.66 MB/sec

Now increase the number of messages published:

  1. % nats-bench -np 1 -n 10000000 -ms 16 foo
  2. Starting benchmark [msgs=10000000, msgsize=16, pubs=1, subs=0]
  3. Pub stats: 7,671,570 msgs/sec ~ 117.06 MB/sec

Run a publish/subscribe throughput test

When using both publishers and subscribers, nats-bench reports aggregate, as well as individual publish and subscribe throughput performance.

Let’s look at throughput for a single publisher with a single subscriber:

  1. % nats-bench -np 1 -ns 1 -n 100000 -ms 16 foo

Note that the output shows the aggregate throughput as well as the individual publisher and subscriber performance:

  1. Starting benchmark [msgs=100000, msgsize=16, pubs=1, subs=1]
  2. NATS Pub/Sub stats: 2,009,230 msgs/sec ~ 30.66 MB/sec
  3.  Pub stats: 1,076,537 msgs/sec ~ 16.43 MB/sec
  4.  Sub stats: 1,004,615 msgs/sec ~ 15.33 MB/sec

Run a 1:N throughput test

When specifying multiple publishers, or multiple subscribers, nats-bench will also report statistics for each publisher and subscriber individually, along with min/max/avg and standard deviation.

Let’s increase both the number of messages, and the number of subscribers.:

  1. % nats-bench -np 1 -ns 5 -n 10000000 -ms 16 foo

Output:

  1. Starting benchmark [msgs=10000000, msgsize=16, pubs=1, subs=5]
  2. NATS Pub/Sub stats: 5,730,851 msgs/sec ~ 87.45 MB/sec
  3.  Pub stats: 955,279 msgs/sec ~ 14.58 MB/sec
  4.  Sub stats: 4,775,709 msgs/sec ~ 72.87 MB/sec
  5.   [1] 955,157 msgs/sec ~ 14.57 MB/sec (10000000 msgs)
  6.   [2] 955,150 msgs/sec ~ 14.57 MB/sec (10000000 msgs)
  7.   [3] 955,157 msgs/sec ~ 14.57 MB/sec (10000000 msgs)
  8.   [4] 955,156 msgs/sec ~ 14.57 MB/sec (10000000 msgs)
  9.   [5] 955,153 msgs/sec ~ 14.57 MB/sec (10000000 msgs)
  10.   min 955,150 | avg 955,154 | max 955,157 | stddev 2 msgs

Run a N:M throughput test

When more than 1 publisher is specified, nats-bench evenly distributes the total number of messages (-n) across the number of publishers (-np).

Now let’s increase the number of publishers and examine the output:

  1. % nats-bench -np 5 -ns 5 -n 10000000 -ms 16 foo

The output:

  1. Starting benchmark [msgs=10000000, msgsize=16, pubs=5, subs=5]
  2. NATS Pub/Sub stats: 6,716,465 msgs/sec ~ 102.49 MB/sec
  3.  Pub stats: 1,119,653 msgs/sec ~ 17.08 MB/sec
  4.   [1] 226,395 msgs/sec ~ 3.45 MB/sec (2000000 msgs)
  5.   [2] 225,955 msgs/sec ~ 3.45 MB/sec (2000000 msgs)
  6.   [3] 225,889 msgs/sec ~ 3.45 MB/sec (2000000 msgs)
  7.   [4] 224,552 msgs/sec ~ 3.43 MB/sec (2000000 msgs)
  8.   [5] 223,933 msgs/sec ~ 3.42 MB/sec (2000000 msgs)
  9.   min 223,933 | avg 225,344 | max 226,395 | stddev 937 msgs
  10.  Sub stats: 5,597,054 msgs/sec ~ 85.40 MB/sec
  11.   [1] 1,119,461 msgs/sec ~ 17.08 MB/sec (10000000 msgs)
  12.   [2] 1,119,466 msgs/sec ~ 17.08 MB/sec (10000000 msgs)
  13.   [3] 1,119,444 msgs/sec ~ 17.08 MB/sec (10000000 msgs)
  14.   [4] 1,119,444 msgs/sec ~ 17.08 MB/sec (10000000 msgs)
  15.   [5] 1,119,430 msgs/sec ~ 17.08 MB/sec (10000000 msgs)
  16.   min 1,119,430 | avg 1,119,449 | max 1,119,466 | stddev 12 msgs