Hello everyone – welcome to the fifth post in our series on Quarkus! (By the way, check out our webinar . We will show how to get started "from scratch" or migrate existing solutions)
In we explored modernizing Java applications using supported Quarkus technologies (CDI and Servlet 3) with the helloworld program from the repository . Today, we will continue with the topic of modernization and discuss memory consumption.
Measuring performance is a fundamental aspect of almost any modernization process, and memory usage reports are an important part of performance analysis. Today, we'll look at relevant measurement tools that can quantitatively evaluate the improvements gained from modernizing Java applications.
Details on measuring memory usage can be found in the Quarkus guide titled
Below, we will simply show how to compare memory usage data across three different application types (JBoss EAP, JAR package, and executable), collecting this data in Linux using the pmap and ps utilities.
JBoss EAP
We run an instance of the JBoss EAP application (see "Deploying helloworld" in ) and then check its process ID (PID) using the following command:
$ pgrep -lf jboss
7268 java
Note. The -a option allows you to retrieve the full command line (i.e.: $ pgrep -af jboss).
Now we use PID 7268 in the ps and pmap commands.
Like this:
$ ps -o pid,rss,command -p 7268
PID RSS COMMAND
7268 665348 java -D[Standalone] -server -verbose:gc -Xloggc:/home/mrizzi/Tools/jboss-eap-7.2.0/jboss-eap-7.2/standalone/log/gc.log -XX:+PrintGCDetails -XX:+PrintGCDateStamps -XX:+UseGCLogFileRotation -XX:NumberOfGCLogFiles=5 -XX:GCLogFileSize=3M -XX:-TraceClassUnloading -Xms1303m -Xmx1303m -XX:MetaspaceSize=96M -XX:MaxMetaspaceSize=256m -Djava.net.preferI
And it looks like this:
$ pmap -x 7268
7268: java -D[Standalone] -server -verbose:gc -Xloggc:/home/mrizzi/Tools/jboss-eap-7.2.0/jboss-eap-7.2/standalone/log/gc.log -XX:+PrintGCDetails -XX:+PrintGCDateStamps -XX:+UseGCLogFileRotation -XX:NumberOfGCLogFiles=5 -XX:GCLogFileSize=3M -XX:-TraceClassUnloading -Xms1303m -Xmx1303m -XX:MetaspaceSize=96M -XX:MaxMetaspaceSize=256m -Djava.net.preferIPv4Stack=true -Djboss.modules.system.pkgs=org.jboss.byteman -Djava.awt.headless=true -Dorg.jboss.boot.log.file=/home/mrizzi/Tools/jboss-eap-7.2.0/jboss-eap-7.2/standa
Address Kbytes RSS Dirty Mode Mapping
00000000ae800000 1348608 435704 435704 rw--- [ anon ]
0000000100d00000 1035264 0 0 ----- [ anon ]
000055e4d2c2f000 4 4 0 r---- java
000055e4d2c30000 4 4 0 r-x-- java
000055e4d2c31000 4 0 0 r---- java
000055e4d2c32000 4 4 4 r---- java
000055e4d2c33000 4 4 4 rw--- java
[...]
ffffffffff600000 4 0 0 r-x-- [ anon ]
---------------- ------- ------- -------
total kB 3263224 672772 643024
We check the RSS value and see that JBoss EAP consumes about 650 MB of memory.
JAR package
We launch the JAR application (see the section 'Running helloworld packaged in JAR' in ):
$ java -jar ./target/helloworld--runner.jar
We again look up the PID using the pgrep command (this time using the described option -a):
$ pgrep -af helloworld
6408 java -jar ./target/helloworld--runner.jar
We run ps and pmap to measure memory usage, but now for process 6408.
Like this:
$ ps -o pid,rss,command -p 6408
PID RSS COMMAND
6408 125732 java -jar ./target/helloworld-quarkus-runner.jar
And it looks like this:
$ pmap -x 6408
6408: java -jar ./target/helloworld-quarkus-runner.jar
Address Kbytes RSS Dirty Mode Mapping
00000005d3200000 337408 0 0 rw--- [ anon ]
00000005e7b80000 5046272 0 0 ----- [ anon ]
000000071bb80000 168448 57576 57576 rw--- [ anon ]
0000000726000000 2523136 0 0 ----- [ anon ]
00000007c0000000 2176 2088 2088 rw--- [ anon ]
00000007c0220000 1046400 0 0 ----- [ anon ]
00005645b85d6000 4 4 0 r---- java
00005645b85d7000 4 4 0 r-x-- java
00005645b85d8000 4 0 0 r---- java
00005645b85d9000 4 4 4 r---- java
00005645b85da000 4 4 4 rw--- java
[...]
ffffffffff600000 4 0 0 r-x-- [ anon ]
---------------- ------- ------- -------
total kB 12421844 133784 115692
We again check the RSS and see that the JAR package consumes about 130 MB.
Executable file
We run the native executable (see the section 'Running the native executable helloworld' in ):
$ ./target/helloworld--runner
We again check its PID:
$ pgrep -af helloworld
6948 ./target/helloworld--runner
And then we use the obtained process identifier (6948) in the ps and pmap commands.
Like this:
$ ps -o pid,rss,command -p 6948
PID RSS COMMAND
6948 19084 .\/target\/helloworld-quarkus-runner
And like this:
$ pmap -x 6948
6948: .\/target\/helloworld-quarkus-runner
Address Kbytes RSS Dirty Mode Mapping
0000000000400000 12 12 0 r---- helloworld-quarkus-runner
0000000000403000 10736 8368 0 r-x-- helloworld-quarkus-runner
0000000000e7f000 7812 6144 0 r---- helloworld-quarkus-runner
0000000001620000 2024 1448 308 rw--- helloworld-quarkus-runner
000000000181a000 4 4 4 r---- helloworld-quarkus-runner
000000000181b000 16 16 12 rw--- helloworld-quarkus-runner
0000000001e10000 1740 156 156 rw--- [ anon ]
[...]
ffffffffff600000 4 0 0 r-x-- [ anon ]
---------------- ------- ------- -------
total kB 1456800 20592 2684
We look at the RSS and see that the executable file occupies about 20 MB of memory.
Comparing memory consumption
So, we got the following figures for memory usage:
- JBoss EAP – 650 MB.
- JAR package – 130 MB.
- Executable file – 20 MB.
Clearly, the executable file takes up much less memory.
Summarizing posts 4 and 5
In this post and the previous one, we discussed the modernization of Java applications using Quarkus-supported technologies (CDI and Servlet 3), as well as various ways to develop, build, and run such applications. We demonstrated how to collect data on memory usage to assess improvements achieved through such modernization. These articles help in understanding how Quarkus works and its benefits—regardless of whether it's about the simplest helloworld program from our examples or much more complex applications from real life.
We will return in two weeks with the final post about Quarkus—see you then!
In our final post, we will show how to combine AMQ Online and Quarkus to build a modern messaging system based on OpenShift using two new technologies related to message processing. Stay tuned for more. .
Source: habr.com
