RemoraJ Stress Test Report

running jBoss 7.2 Remora 0.1.4-SNAPSHOT built 2020-02-28 and RemoraJ 0.1.9 built 2020-07-09 With testHarnesses built 2020-02-28:

	Threads	Schedulle (every ..)	Target
ApacheHttpClientHarness	50	50ms	Same jBoss
SQLHarness	50	50ms	SQL server running the same machine
MQReceiveHarness	50	50ms	MQ running on VirtualBox, same queue as send
MQSendHarness	50	50ms	MQ running on VirtualBox, same queue as receive
WebsocketSendHarness	50	50ms	Same jBoss, testHarnness endpoint
KafkaConsumerHarness	50	50ms	Kafka running on the same machine
KafkaProducerHarness	50	50ms	Kafka running on the same machine

Measuring with JConsole.

Results:

With RemoraJ 0.1.4-SNAPSHOT

Heap Memory Usage: ~900 CPU usage: 50-60

img/image1.png

img/image2.png

With RemoraJ 0.1.9

Heap Memory Usage: ~450 CPU usage: <10%

img/image1.1.9.png

img/image2.1.9.png

With RemoraJ without chronicle queue

The same test condition except chronicle queue opted out, no files are created or written.

Heap Memory Usage: ~900 CPU usage: 50-55%

img/image3.png

img/image4.png

Without RemoraJ

-javaagent option is commented out;

Heap Memory Usage: ~300 CPU usage: <10%

img/image5.png

Conclusions

For test scenario where ~350 thread are running invoking various instrumented remote services (HTTP, JMS send and receive) scenario RemoraJ causes overhead as expected. This might be caused by GC as jBoss was running with „JAVA_OPTS=-Xms1G -Xmx1G -XX:MetaspaceSize=96M -XX:MaxMetaspaceSize=256m“. Further investigation needed.

|