JVM Camel Analisys

The objective of this report is compare memory usage by Apache Camel in a simple route, this simple route expose a CRUD rest operation with 50 objects persisted in a IMDB in different scenarios.

• Camel with Spring Boot

• Camel with Jetty

• Vertx (no camel) - I tested with Vertx to compare the memory consumption with a non camel application.

All data was collected with 10 request per second of load.

Execution flow

1. Run the application

2. Run the load test

3. Collect the metrics

Report Summary

You could visualize this spread sheet here.

Result with no tuning

Result with MaxRAM=100m

Result with MaxRAM=300m

Hardware spec

The test was did using one Macbook Pro (Retina, 13-inch, Late 2013):

macOS Sierra (Version 10.12.6)
Processor 2,8 GHz Intel Core i7
Memory 16 GB 1600 MHz DDR3

• Java version "1.8.0_131"

• Apache Maven 3.5.0

Fuse Integration Services (Spring Boot)

Project: jvm-analisys/fis-rest

No tuning

export MAVEN_OPTS='-XX:NativeMemoryTracking=summary'; mvn spring-boot:run

Running with MaxRAM=100m

export MAVEN_OPTS='-XX:NativeMemoryTracking=summary -XX:MaxRAM=100m'; mvn spring-boot:run

Running with MaxRAM=200m

export MAVEN_OPTS='-XX:NativeMemoryTracking=summary -XX:MaxRAM=200m'; mvn spring-boot:run

Camel with Jetty

Project: jvm-analisys/camel-examples/camel-example-restlet-jdbc

load database

for i in {1..50}
do
      curl -X POST -d "firstName=test&lastName=person" http://localhost:8080/rs/persons;
done

No tuning

export MAVEN_OPTS="-XX:NativeMemoryTracking=summary"; mvn jetty:run -Dimpl=java-rest-dsl

Running with MaxRAM=100m

export MAVEN_OPTS="-XX:NativeMemoryTracking=summary -XX:MaxRAM=100m"; mvn jetty:run -Dimpl=java-rest-dsl

Result: GC overhead limit exceeded.

Running with MaxRAM=200m

export MAVEN_OPTS="-XX:NativeMemoryTracking=summary -XX:MaxRAM=100m"; mvn jetty:run -Dimpl=java-rest-dsl

Vertx

The sample used in vertx is the following: /web-examples/src/main/java/io/vertx/example/web/rest/SimpleREST.java

I ran the class direct by the IDE (main method).

No tuning

Running with MaxRAM=100m

Running with MaxRAM=200m

Load Test

The testing tool used was Locust. All data was collected with a 10 request per second load.

To run the load test use:

locust -f <project-path>/load-test.py --host=http://localhost:8080

After that access the web console: http://localhost:8089/

I used 100 users with 1 of hatch rate.

Collect metrics

The metrics was collected using Java Memory Tracking jdk tool with the option summary.

To simplify the collect proccess I developed this tool https://github.com/hodrigohamalho/jmt-analyzer