Skip to content

Latest commit

 

History

History
541 lines (351 loc) · 25.1 KB

MICRONAUT-REACTIVE_VS_MICRONAUT-REACTIVE-NATIVE.md

File metadata and controls

541 lines (351 loc) · 25.1 KB
Raw

CHAPTER 1: IF YOU’RE GOOD AT SOMETHING, NEVER DO IT FOR FREE.

Hello, world,

Today I would like to test Micronaut:

A MODERN, JVM-BASED, FULL-STACK FRAMEWORK FOR BUILDING MODULAR, EASILY TESTABLE MICROSERVICE AND SERVERLESS APPLICATIONS.

Modern... talking? :) Let's check it.

Micronaut says that it's faster and less expensive due to the fact of lower resources consumption and better serialization during the build time.

Trust but verify. So, I am going to verify it.

I will test the most efficient approach of Spring and Quarkus to compare it. And I would like to test the reactive approach using reactor-http and data-r2dbc with r2dc-pool and vertx-pg-client (same to Quarkus, yeah!).

I hope you are familiar with my previous performance results regarding the Spring Web (as Native), and Spring Reactive (as Native), and Quarkus Reactive (as Native). If not, please take a look on it:

This article is not about Micronaut internal architecture and design, its paradigms, and the solutions that Micronaut Team brings to life. This article is about performance.

I will not dive deeper into the Micronaut Reactive stack and the business description of my application you could always read it on your own in my previous research and in the documentations.

And today I will check the performance of a native executable (including in the docker solution) and a default one (including inD solution as well).

CHAPTER 2: WHAT DOESN'T KILL YOU, SIMPLY MAKES YOU STRONGER!

So, we are going to create our application based on Micronaut. Hopefully, the developers who are familiar with Spring or Quarkus could easily migrate to Micronaut. But... I've faced several problems during the migration:

  1. Micronaut doesn't support Money type in Postgres. And that's cool;
  2. Not so clear documentation as it is for Quarkus and Spring. In some cases it's really poor;
  3. To avoid workarounds on the backend part with micronaut data lib, I had to rename several columns i.e. origin -> origin_id;
  4. If not to define connections as 'pool' you are working with connection per thread model;
  5. Doesn't support Gradle 8.0.1. The team doesn't follow one of the major building tools. Didn't prepare using RC of Gradle.

The languages, frameworks, and tools I used:

JDK GC Gradle Micronaut Gatling Dive Grype
17 G1 8.0.1 3.8.5 3.9.2 0.10.0 0.57.1

My local setup:

Model Processor Processor Speed Number of Processors Cores L2 Cache (per Core) L3 Cache Memory
MacBook Pro Quad-Core Intel Core i7 2,3 GHz 1 4 512KB 8MB 32GB

I will highlight some of the configurations here.

Gradle Build Script

// https://github.com/bmuschko/gradle-docker-plugin/issues/1035
buildscript {
    dependencies {
        classpath("com.github.docker-java:docker-java:3.3.0")
        classpath("com.github.docker-java:docker-java-transport-httpclient5:3.3.0")
    }
}

plugins {
    id("com.github.johnrengelman.shadow") version "7.1.2"
    id("com.google.cloud.tools.jib") version "3.3.1"
    id("io.micronaut.minimal.application") version "3.7.2"
    id("io.micronaut.graalvm") version "3.7.2"
    id("io.micronaut.docker") version "3.7.2"
    id("io.micronaut.aot") version "3.7.2"
    id("java")
}

version = "0.0.1-SNAPSHOT"
group = "by.vk"

application {
    mainClass.set("by.vk.Application")
}

java {
    sourceCompatibility = JavaVersion.toVersion("17")
    targetCompatibility = JavaVersion.toVersion("17")
}

repositories {
    mavenCentral()
}

dependencies {
    annotationProcessor("io.micronaut.data:micronaut-data-processor")
    annotationProcessor("io.micronaut.serde:micronaut-serde-processor")
    implementation("io.micronaut.serde:micronaut-serde-jackson")
    implementation("io.micronaut.data:micronaut-data-r2dbc")
    implementation("io.micronaut.reactor:micronaut-reactor-http-client")
    implementation("jakarta.persistence:jakarta.persistence-api:3.1.0")
    runtimeOnly("ch.qos.logback:logback-classic")
    runtimeOnly("io.vertx:vertx-pg-client")
    runtimeOnly("io.r2dbc:r2dbc-pool")
    runtimeOnly("org.postgresql:r2dbc-postgresql:1.0.1.RELEASE")
}

tasks {
    jib {
        to {
            image = 'micronaut-reactive-distroless:latest'
        }
        from {
            image = "gcr.io/distroless/java17"
        }
        container {
            jvmFlags = ['-noverify', '-XX:+UseContainerSupport', '-XX:MaxRAMPercentage=75.0', '-XX:InitialRAMPercentage=50.0', '-XX:+OptimizeStringConcat', '-XX:+UseStringDeduplication', '-XX:+ExitOnOutOfMemoryError', '-XX:+AlwaysActAsServerClassMachine', '-Xmx512m', '-Xms128m', '-XX:MaxMetaspaceSize=128m', '-XX:MaxDirectMemorySize=256m', '-XX:+HeapDumpOnOutOfMemoryError', '-XX:HeapDumpPath=/opt/tmp/heapdump.bin', '-Djava.rmi.server.hostname=localhost', '-Dcom.sun.management.jmxremote=true', '-Dcom.sun.management.jmxremote.rmi.port=8051','-Dcom.sun.management.jmxremote.port=8051', '-Dcom.sun.management.jmxremote.local.only=false', '-Dcom.sun.management.jmxremote.authenticate=false', '-Dcom.sun.management.jmxremote.ssl=false']
            ports = ['8080', '8051']
            labels.set([maintainer: 'Vadzim Kavalkou <[email protected]>', appname: 'a2b-service', version: '0.0.1-SNAPSHOT'])
            creationTime.set('USE_CURRENT_TIMESTAMP')
        }
    }
    dockerfile {
        exposedPorts.set([8080, 8051])
        args = ['-XX:+UseContainerSupport', '-XX:MaxRAMPercentage=75.0', '-XX:InitialRAMPercentage=50.0', '-XX:+OptimizeStringConcat', '-XX:+UseStringDeduplication', '-XX:+ExitOnOutOfMemoryError', '-XX:+AlwaysActAsServerClassMachine', '-Xmx512m', '-Xms128m', '-XX:MaxMetaspaceSize=128m', '-XX:MaxDirectMemorySize=256m', '-XX:+HeapDumpOnOutOfMemoryError', '-XX:HeapDumpPath=/opt/tmp/heapdump.bin', '-Djava.rmi.server.hostname=localhost', '-Dcom.sun.management.jmxremote=true', '-Dcom.sun.management.jmxremote.rmi.port=8051','-Dcom.sun.management.jmxremote.port=8051', '-Dcom.sun.management.jmxremote.local.only=false', '-Dcom.sun.management.jmxremote.authenticate=false', '-Dcom.sun.management.jmxremote.ssl=false']
    }
}

micronaut {
    runtime("netty")
    processing {
        incremental(true)
        annotations("by.vk.*")
    }
    aot {
        cacheEnvironment = true
        optimizeServiceLoading = true
        optimizeClassLoading = true
        convertYamlToJava = true
        precomputeOperations = true
        deduceEnvironment = true
        optimizeNetty = true
    }

}

And application.yml, for sure.

micronaut:
  application:
    name: micronaut-reactive
  server:
    port: 8080
    context-path: "/api/v1"
    netty:
      responses:
        file:
          cache-seconds: 0

netty:
  default:
    allocator:
      use-cache-for-all-threads: false

# https://micronaut-projects.github.io/micronaut-sql/latest/guide/configurationreference.html
# https://github.com/eclipse-vertx/vert.x/blob/master/src/main/java/io/vertx/core/VertxOptions.java#L38
r2dbc:
  datasources:
    options:
      useDaemonThread: true
      driver: pool
      protocol: postgres
      host: postgres-a2b
      port: 5432
      username: postgres
      password: postgres
      database: a2b
      connectTimeout: 35000ms

As I've already mentioned, I decided to check all possible types of launching the application, such as jar, jar in docker, native executable, and in docker native solutions as well.

In Micronaut you have several ways of building a docker image:

  • JAR in Docker;
  • JIB;
  • GraalVM native executable in Docker;
  • CRaC;
  • AOT optimized docker image.

I will provide you the link to sources in the end of this article.

And now "Just do it".

CHAPTER 3: THEY LAUGH AT ME BECAUSE I'M DIFFERENT. I LAUGH AT THEN BECAUSE THEY'RE ALL THE SAME.

There are results of non-native solutions.

  • JAR

Global information:

Requests:

Requests per second:

Responses per second:

Response time for first minute:

Response time for all time:

You could download the JAR Performance Tests Results and check it on your own.

  • JIB with Distroless base image

Global information:

Requests:

Requests per second:

Responses per second:

Response time for first minute:

Response time for all time:

Docker image investigation:

Dive:

Grype:

You could download the Distroless Performance Tests Results and check it on your own.

  • Default Docker Image

Global information:

Requests:

Requests per second:

Responses per second:

Response time for first minute:

Response time for all time:

Docker image investigation:

Dive:

Grype:

You could download the Default Docker Image Performance Tests Results and check it on your own.

Let's gather all the information:

TYPE BUILD TIME (s) ARTIFACT SIZE (MB) BOOT UP (s) RPS RESPONSE TIME (95th pct) (ms) SATURATION POINT RAM (MB) CPU (%) THREADS (MAX) POSTGRES CPU (%)
JAR 17 21.4 1.176 616.132 19540 2487 4495 10 49 75
DEFAULT DOCKER IMAGE 74 346 2.373 425.895 41730 1189 586 59 23 26
JIB 21 252 1.767 416.672 37474 1023 586 60 26 31

Move on.

CHAPTER 4: IF YOU’RE GOOD AT SOMETHING, NEVER DO IT FOR FREE.

Now it's a time to compare previous solutions with native ones.

  • Native Executable

Global information:

Requests:

Requests per second:

Responses per second:

Response time for first minute:

Response time for all time:

You could download the Native Executable Performance Tests Results and check it on your own.

  • Default Native Docker Image

Firstly, I will demonstrate you errors.

Checksums ".n.h.c.c.DecompressionException: CRC value mismatch." Checksums fail everywhere. And logs are clear.

Global information:

Requests:

Requests per second:

Responses per second:

Response time for first minute:

Response time for all time:

Docker image investigation:

Dive:

Grype:

You could download the Native Executable in Docker Performance Tests Results and check it on your own.

  • AOT Optimized Docker Image

Again, the errors.

Checksums ".n.h.c.c.DecompressionException: CRC value mismatch." Checksums fail everywhere. And logs are clear.

Global information:

Requests:

Requests per second:

Responses per second:

Response time for first minute:

Response time for all time:

Docker image investigation:

Dive:

Grype:

You could download the AOT Optimized Docker Image Performance Tests Results and check it on your own.

TYPE BUILD TIME (s) ARTIFACT SIZE (MB) BOOT UP (s) RPS RESPONSE TIME (95th pct) (ms) SATURATION POINT RAM (MB) CPU (%) THREADS (MAX) POSTGRES CPU (%)
NATIVE EXECUTABLE 188 78 0.042 534.58 25728 1717 762 38 35 65
DEFAULT NATIVE DOCKER IMAGE 339 99.3 0.115 338.455 46090 1072 419.2 50 14 25
AOT OPTIMIZED DOCKER IMAGE 5808 99.4 0.088 308.204 53506 1095 381 60 15 12
CHAPTER 6: WHY SO SERIOUS?

Let's compare all the results including the Spring Web, Spring Reactive, Quarkus, and their native solutions as well.

FRAMEWORK APPLICATION TYPE BUILD TYPE BUILD TIME (s) ARTIFACT SIZE (MB) BOOT UP (s) TOTAL REQUESTS KO(%) RPS RESPONSE TIME (95th pct) (ms) SATURATION POINT RAM (MB) CPU (%) THREADS (MAX) POSTGRES CPU (%)
SPRING WEB NATIVE BUILD PACK 751 144.79 1,585 453012 25 374.566 47831 584 310 12.5 64 99
NATIVE BUILD TOOLS 210 116.20 0.310 480763 29 414.785 32175 1829 263 8 52 99
UNDERTOW 5 49.70 3.59 523756 24 381.127 50977 1611 658 11 33 99
UNDERTOW IN DOCKER 46 280 5.20 430673 33 448.682 29998 916 840 15 32 99
REACTIVE + R2DBC NATIVE BUILD PACK 1243 98.5 0.103 691487 17 615.750 17891 1904 685 30 14 70
NATIVE BUILD TOOLS 187 71.7 0,107 1013549 10 934.147 12591 3038 634 32 23 70
JAR 3.1 40.6 2.55 1168782 8 1091.30 10406 4391 1823 8 31 70
JAR IN DOCKER 39 271 3.95 699180 17 631.599 18955 2250 883 29 31 70
QUARKUS REACTIVE + R2DBC FAST JAR 4 N/A 0.987 828711 13 755.434 13686 1971 1054 9 25 99
UBER JAR 8 17.7 1.884 826311 13 753.933 14111 2149 989 5 23 99
JIB WITH UBI 16 384 1.151 661502 18 593.275 20170 1305 1054 8 26 70
JIB WITH DISTROLESS 14 249 1.088 473991 20 540.492 33060 1339 970 8 26 93
DOCKER 39 416 0.948 609675 28 428.563 24206 1315 262 18 21 53
NATIVE EXECUTABLE 180 49.3 0.223 768017 15 697.563 16426 1967 646 10 15 99
+ UPX-MAX NATIVE EXECUTABLE 741 15 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A
NATIVE MICRO BASE IMAGE 301 78.6 0.031 570959 22 507.971 25637 1282 690 20 8 57
NATIVE MINIMAL BASE IMAGE 301 152 0.025 523534 25 448.231 35777 914 669 17 8 61
NATIVE DISTROLESS BASE IMAGE * 238 72.1 0.032 546371 23 473.458 30156 1747 622 23 8 45
NATIVE DISTROLESS BASE IMAGE * + ** 238 72.1 0.037 584874 21 515.762 23786 2254 628 17 8 47
MICRONAUT REACTIVE + R2DBC JAR 17 21.4 1.176 683907 17 616.132 19540 2487 4495 10 49 75
DEFAULT DOCKER IMAGE 74 346 2.373 488076 27 425.895 41730 1189 586 59 23 26
JIB 21 252 1.767 489590 27 416.672 37474 1023 586 26 60 31
NATIVE EXECUTABLE 188 78 0.042 604610 20 534.580 25728 1717 762 38 35 65
DEFAULT NATIVE DOCKER IMAGE 457 99.3 0.115 371624 44 338.455 46090 1072 419 50 14 25
AOT OPTIMIZED DOCKER IMAGE 5808 99.4 0.088 339333 49 308.204 53506 1095 381 60 15 12

ACTIVE USERS ~10k

* is experimental feature;

** with --security-opt seccomp=unconfined and volume creation.

A bit of magic and the charts appear. They include the data of Spring Web, Spring Reactive, Quarkus Reactive, Micronaut Reactive and their native suggestions.

  • Let's compare basic solutions that provides us with JARS after the build:

  • JAR IN DOCKER:

  • NATIVES:

  • NATIVE IN DOCKER:

Actually, I could share my thoughts about Micronaut and compare it with Reactive solutions in Spring and Quarkus:

  • Not so good and clear documentation as for Spring and Quarkus;
  • Some of these approaches don't work without workarounds;
  • Doesn't support the newest MAJOR Gradle releases;
  • Native solutions has less vulnerabilities. But ...
  • Native in Docker solution doesn't work properly with netty. Checksums ".n.h.c.c.DecompressionException: CRC value mismatch." fail everywhere. And logs are clear.

What to bring into production is up to you. Few years ago it was a game changer in the microframeworks area but not today. This framework is not for me. Too much effort to solve the problems that should be done out of the box nowadays.

PS: AS YOU KNOW, MADNESS IS LIKE GRAVITY...ALL IT TAKES IS A LITTLE PUSH.

This article is the 4th in my performance journey.

Next, I will bring you details regarding the Vert.x, Helidon, and Ktor.

So, will be in touch.

HAVE A NICE DAY.

SOURCES

BACK TO THE MAIN PAGE