Suggest e.g. a class
performance/test/perf/src/org/netbeans/performance/jars/JarTest.java
with contents something like this:

public class JarTest extends Benchmark {
    public static void main(String[] args) {
        simpleRun(JarTest.class);
    }
    public JarTest(String name) {
        super(name, new int[][] {
            // {numberOfJars, numberOfFilesPerJar}
            new int[] {5, 5},
            new int[] {100, 5},
            new int[] {5, 100},
            new int[] {100, 100},
        });
    }
    protected void setUp() throws Exception {
        int[] arg = (int[])getArgument();
        int numberOfJars = arg[0];
        int filesPerJar = arg[1];
        // XXX
    }
    protected void tearDown() throws Exception {
        // XXX
    }
    public void testXXX() throws Exception {
        int count = getIterationCount();
        int[] arg = (int[])getArgument();
        int numberOfJars = arg[0];
        int filesPerJar = arg[1];
        for (int i = 0; i < count; i++) {
            // XXX do something
        }
    }
}

I've somehow measured the performance of module system using a lot
of jars (1000), it covers opening 2xmodulecount jar files.
See the comments by the issue 21675.

Target milestone was changed from '3.4' to TBD.

Target milestone was changed from '3.4' to TBD.

passing to Petr for further measurements, as Jesse suggested.
Marking as STARTED because we have something already.

Reassigning changes STARTED to NEW.

Charter of the general performance team, I think...

Petr can have more uptodate data now as he is rewriting our classloading to better handle opening of JAR files.

Probably now obsolete.