Java Performance: Ubuntu Linux vs. Windows Vista

[kraftman]

Oh my, it's two years old article...

[kraftman]

Linux got owned where it matters the most: Graphics.

I don't agree. It was owned in this test not as much as someone would suppose (according to graphic drivers state on Linux), but I consider non graphic related benchmarks (where Linux simply owned Windows) are far more important. Java graphic performance doesn't affect so many things then non graphic java performance.

[lasner12]

A number big players have come up with pages or applications in an attempt to draw World Cup fans into either combining using a laptop whilst watching the live games on television or using the application as their main view into the World Cup. The problem is that nobody has got it quite right. With a live event the data that you see in your World Cup web application needs to be correct with up to the second information or it loses its value. None of the big players have managed this. The data also needs to be relevant and of good quality. The Twitter application fails in this respect.

[akpch]

Interesting results, though it's not clear to me whether the Java performance in Linux is better because the Linux Java engine is better, or because system resource overhead to run Ubuntu is different (lower maybe?) than that to run Windows (another form of the OS 'tax').

I'm assuming that there are no major resource-draining processes running in the background, but it would be nice to see the results of running the Java benchmarks inside VMs for comparison to the native OS results; that is, in addition to the current methodology, run the tests inside an Ubuntu VM running inside the Windows OS and/or a Windows VM running inside the Ubuntu OS. In fact, it might be possible to run a native and a VM Java benchmark AT THE SAME TIME, which would resolve my questions about whether system overhead is an important factor influencing the results.

Figured strange results eventually. I was using sin and cos functions in my benchmark which are traced down to native routines. Once switched to something more neutral (xml parsing, string processing) results are very close. And of course in file operations with multiple small files and ext4 Linux is way ahead.

[fxfuji]

Interesting results, though it's not clear to me whether the Java performance in Linux is better because the Linux Java engine is better, or because system resource overhead to run Ubuntu is different (lower maybe?) than that to run Windows (another form of the OS 'tax').

I'm assuming that there are no major resource-draining processes running in the background, but it would be nice to see the results of running the Java benchmarks inside VMs for comparison to the native OS results; that is, in addition to the current methodology, run the tests inside an Ubuntu VM running inside the Windows OS and/or a Windows VM running inside the Ubuntu OS. In fact, it might be possible to run a native and a VM Java benchmark AT THE SAME TIME, which would resolve my questions about whether system overhead is an important factor influencing the results.

[kraftman]

No, if I want to comapre OSes, not JREs I will pick whatewer RUNS FASTER on the SPECIFIC PLATFORM.

I admit, should have tried JRockit as well.

So, you should pick WHATEVER on EACH platform to know what runs faster on specific platform.

[akpch]

If you want to compare OS'es not different versions of java use the SAME java on each OS...

No, if I want to comapre OSes, not JREs I will pick whatewer RUNS FASTER on the SPECIFIC PLATFORM.

I admit, should have tried JRockit as well.

[kraftman]

I did some testing on my own. Having run tests with -server switch on windows shows great improvement. The test was multithreaded heavy math calculations.

Linux was latest kernel from Ubuntu 9.04 (tried both server and desktop flavor) and best result on Vista using SUN Java 6_13 was 15% better than best result on Linux using Open JVM.

Solaris on the other hand was about 25% faster using 64 bit java 1.5

Regards
Alex

If you want to compare OS'es not different versions of java use the SAME java on each OS...

[akpch]

I did some testing on my own. Having run tests with -server switch on windows shows great improvement. The test was multithreaded heavy math calculations.

Linux was latest kernel from Ubuntu 9.04 (tried both server and desktop flavor) and best result on Vista using SUN Java 6_13 was 15% better than best result on Linux using Open JVM.

Solaris on the other hand was about 25% faster using 64 bit java 1.5

Regards
Alex

[cayhorstmann]

Questions about benchmark

I posted this in the JavaChampions mailing list, and got these questions:

"I wonder why he used JDK 1.6.0_10 on Ubuntu and 1.6.0_07 on Windows. There has been enough work on the _10 JIT that it's not a fair comparison.

Also, he does not say whether he is using the -server of -client hotspot compilers for these tests. It could be that Ubuntu defaulted to -server and Windows to -client, which would explain the results..."

I think these are fair questions, and I'd like to know the answers so I can report back.

Thanks,

Cay