How different OS/kernel affect model performance

You know something useful, tips&tricks, have some unofficial documentation...? Please share your knowledge with us in this section of EMS forum. * * * Please DON'T open topics with questions in this section! * * *
Post Reply
Posts: 1604
Joined: Wed Aug 19, 2009 10:05 am

How different OS/kernel affect model performance

Post by meteoadriatic »

Long time ago I was curious about this; does operating system or kernel version affect WRF performance or not?

I'm not especially knowledgeable on how kernel works with hardware and can better kernel squeeze any more performance from CPU, so I did a test to answer my doubt!

I first installed Centos 5.8, latest updates and run my simulation, then wiped everything from disk, installed Centos 6.2, latest updates, rsynched back /usr1 and /home partition and run the same simulation again.

This is not official benchmark case but my own working domain. Results are:
Centos 5.8 -> 29:43
Centos 6.2 -> 29:32

I did just this one test. Difference is small and, IMHO, can be assigned to statistical error. But again, it might come from better kernel or other part of system. I don't know. You decide. :mrgreen:

Other small difference on test systems is filesystem type. Centos 5.8 had ext3, 6.2 has ext4 formatted partitions. I don't think this affect model performance at all, because there isn't much I/O load. And also everything else in both cases is default Centos (minimal) installation.

Hardware was Core2Quad Q9450, 4GB RAM.

Now, if first kernel was from 5.0 or even Centos 4.x there might be different results, I can't tell. Also, if I used newer hardware, like Ivy Bridge CPU, again I don't know if newer kernels could drive them better than old. In next month I expect to purchase i7 3770K CPU based system so we will see :)
Post Reply