Question

I’m executing a Java program 10,000 times in Eclipse indigo. All operations performed are in memory; no disk IO or network IO. I noticed that over the period of time the response time increased from 2msec to 25 msec for the same input. Upon plotting the response time, the graph looks linear. First thought was I have not tuned the garbage collector. I have set garbage collector for maximum throughput and less gc pause times. Still no difference.

For screenshots on response time and Java VM GC, please click the link

I’m running Oracle Java on Macbook Pro 2011.

java version "1.6.0_29"
Java(TM) SE Runtime Environment (build 1.6.0_29-b11-402-11M3527)
Java HotSpot(TM) 64-Bit Server VM (build 20.4-b02-402, mixed mode)

Following are my VM settings

-server -XX:+UseParallelGC -XX:+UseParallelOldGC -Xms48m -Xmx64m -XX:NewSize=32m -XX:MaxNewSize=32m -Xss2048k -XX:SurvivorRatio=16 -XX:MaxGCPauseMillis=15 -verbose:gc -Xloggc:gclog.txt -XX:+PrintGCDetails

Appreciate it, if any one could shed light on this.

Answer 1

It is difficult to say what the problem is, based on what you’ve told us.

• It could be a problem in your application per se.

• It could be a problem in the way you are launching your application.

• It could be a problem in the way you are measuring “response time”.

---

If you want a better answer you need to explain how you are “executing a Java program 10,000 times in Eclipse indigo”.

• How / where are the response times being measured? By Eclipse? By something else?
• Does this really mean that you are running the Eclipse launcher 10,000 times?
• Or are you running some benchmarking framework class that is launching the application?
• Or is it just calling the application’s main 10,000 times?
• Is Eclipse running the test in a separate JVM or in the Eclipse JVM?

Also:

• When you say there is no network I/O, does that include 127.0.0.x and local IPC?
• What is the application actually doing? Is it talking to something else in some shape or form?