IOzone filesystem benchmarker - a couple of questions
Hello everyone, I am hoping that maybe someone could help me out with this. I am going to do a project on linux filesytems and NTFS/FAT too.
I have borrowed an acuta accusys 4 disk raid external array from my supervisor for this, which supports levels 0, 0+1, and 5. I can use 4 disks in the raid setting.
I have never really used any tools like IOzone before and am practically clueless. What i'm aiming for here is a fairly straightforward comparison between them which will basically point out that a certain fs performs faster at writing than another, or maybe speculate that a certain fs benefits more from a certain RAID level more than another fs, and hopefully give a reason as to why.
I was hoping that I could get a few suggestions about things to point out.
I'm going to research xfs, jfs,ext2,ext3, and raiserfs. I have no idea how to even make these filesystems :S but thats a problem for later.
I was looking at the strided read test and was wondering, what kind of applications might behave like this? With this external array I have no idea how to set stripe size etc or even an idea of what the chunk size is.
If I were to run the strided read test just once for NTFS, just to show that I made an extra effort, what could I say about the results? Any suggestions on how I could use the RAID array, to perhaps demonstrate some anomalies and then say "This happened cos of xxxxx and this is why people should make sure that their filesystem/RAID is suitable for these kind of applications."
Also, how can you tell from the IOzone results if a filesystem is cpu intensive or not?
As it stands I am hoping to do the basic read/write reread/rewrite, strided read tests.
I will be eternally grateful for any helpful input that could help me talk about the results and not just present a big ol'lonely graph
Thanks, in advance.