/stream-startup

Primary LanguageJava

Benchmarks for Spring AMPQ, Spring Integration and Spring Cloud Stream startup time.

Results

class       method  sample  beans  classes      heap    memory  median  mean  range
StreamBenchmark  main  empt  24.000   3781.000  5.055   38.345  0.520  0.525  0.005
StreamBenchmark  main  amqp  49.000   3974.000  6.443   41.465  0.598  0.607  0.014
StreamBenchmark  main  intg  77.000   4571.000  6.640   45.664  0.759  0.772  0.021
StreamBenchmark  main  jlog  668.000  5256.000  10.471  59.601  1.086  1.098  0.012
StreamBenchmark  main  demo  668.000  5682.000  10.012  61.274  1.133  1.151  0.021

Key:

  • empt - an empty application, only Spring Boot

  • amqp - a Spring Boot application with a Rabbit listener

  • intg - just Spring Integration autoconfiguration

  • jlog - the demo sample with classpath exclusions (validator and logback)

  • demo - Spring Cloud Stream and Rabbit with a stream listener

Running

You can run the benchmarks in the IDE (or on the command line) by running the test class StreamBenchmark. Open it up in the IDE and make sure annotation processing is switched on (update the Maven project and check the compiler settings in the project in Eclipse). Run the test with -DpublishTo=csv:target/result.csv to get the results in a nice CSV format at the end.

If you want to just run one of the samples edit the MainState and add the name of the sample to the @Param annotation, e.g.

		@Param("demo")
		private Sample sample;

runs just the demo sample.

How it Works

The different samples are run with different classpaths to control the Spring Boot autoconfiguration. The classpath is computed using the Thin Launcher using a "profile" whose name is the sample name. So you will find files in META-INF/thin-*.properties corresponding to the sample names. The demo sample is canonical - it runs with the full classpath from the pom.xml.