Windows与Unix运行相同的Java代码时的性能差异

Question

I am running below code via eclipse in windows and then in Unix as standalone java.

import java.io.ByteArrayInputStream;
import java.io.ByteArrayOutputStream;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileNotFoundException;
import java.io.IOException;
import java.io.InputStream;
import java.io.ObjectInputStream;
import java.sql.SQLException;
import java.text.ParseException;

public class readObjectBoeing {



/**
 * @param args
 * @throws ParseException 
 * @throws SQLException 
 */
public static void main(String[] args) {
    //File file = new File("/opt/app/d1ebp1m1/dv01/Vibhor/test/00017741_repository.dat");
    File file = new File("C:/_Vibhor/00017741_repository.dat");
    InputStream is;
    try {
        is = new FileInputStream(file);
        byte[] b = toByteArray(is);//read from file;
        Object o1 =null;
        o1 = convertByteArrayToObject(b);
    } catch (FileNotFoundException e) {
        // TODO Auto-generated catch block
        e.printStackTrace();
    } catch (IOException e) {
        // TODO Auto-generated catch block
        e.printStackTrace();
    } catch (Exception e) {
        // TODO Auto-generated catch block
        e.printStackTrace();
    }

}
public static Object convertByteArrayToObject(byte[] buf) throws Exception 
{

    if (buf.length == 0)
    {
        return null;
    }
    long startTime = -1;
    long step1=-1,step2=-1;
    Object                  obj = null;
    ByteArrayInputStream    bis = null;
    ObjectInputStream       in  = null;
    try 
    {   
        bis = new ByteArrayInputStream(buf);
        in  = new ObjectInputStream(bis);
        startTime = System.currentTimeMillis()/1000;
        obj = in.readObject();
        step1 = System.currentTimeMillis()/1000 - startTime ;
        System.out.println("in.readObject()!! :  " + step1);
    } 
    catch (Exception e) 
    {
        throw e;
    } 
    finally 
    {
        if (in != null) 
        {
            in.close();
        }
        if (bis != null) 
        {
            bis.close();
        }
        in  = null;
        bis = null;
    }

    return obj;
}
public static byte[] toByteArray(InputStream input) throws IOException
{
    ByteArrayOutputStream output = new ByteArrayOutputStream();
    long count = 0L;
    byte[] buffer =new byte[4096];
    for(int n = 0; -1 != (n = input.read(buffer));){
        output.write(buffer, 0, n);
        count += n;
    }
    return output.toByteArray();
}

}

00017741_repository.dat - it is 57Mb file.
In windows obj = in.readObject(); - it takes me 4-5 seconds.
But in Unix obj = in.readObject(); it takes me 19 - 25 sec!

I am using VM args -Xmx512m to execute in both cases.

In unix:

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

In Windows:

jre 1.6.0_26

What am I missing here? Any suggestions to improve the performance in unix?

Answer 1

It is not all together that shocking to find differences like this between platforms. Understand that while the bytecode is the same the JVM is platform specific.

There are a number of areas like this, where you have code that works perfectly on one platform and fails on another. One case I ran into was in drag n drop with files, where unix (at least ubuntu) drops files with a different dataflavor than windows.

You must test on all platforms you intend for your code to run in

Also, in your specific case, you should either try to read from your inputstream more efficiently (using a byte[]) or create a BufferedInputStream to wrap your FileInputStream in

Answer 2

Here are two specific things you can do to figure this one out:

1. make sure that on Unix the file resides on a local file system and not on a network mount;
2. use a profiler (such as YourKit ) to find out where the time is spent.

Answer 3

You also have to differ between different JVMs used. For example on *nix systems there are vms like "icedtea" or "OpenJDK" widely used(because they are the default installation).