I am running a hadoop I/O workload in redhat linux and have ran that in IBM AIX as well with same underlying hardware, disk and disk configuration. With the default settings on the os redhat linux is giving more I/O throuhput compared to IBM AIX.

So how different is the device drivers configurations in redhat linux from IBM AIX for handling I/O from the hard disk ?

I know Redhat Linux has a disk scheduling algorithm namely by default as CFQ(Completely Fair Queue), Noop scheduler, anticipatory scheduler,deadline scheduler. does these also effect the I/O throughput of the disk. and as far as i know AIX does not have these kind of scheduling algorithm. we can increase the depth of the queue in AIX but that is also not increasing the throuhput of the disk much.

The device driver implementation are different in both the cases as i know is the device driver implementation of redhat linux is more efficient in handling the I/O that is why am i seeing this kind of difference ?

Or is it that AIX can be optimized to get more throuhput as that of Redhat Linux ?

  • 1
    Are you running that test on the same hardware (including disks)?
    – Mat
    Sep 28 '13 at 10:46
  • yeah on same hardware i will mention that in the question. Sep 28 '13 at 10:50
  • It might also depend on the specific load pattern.
    – peterph
    Sep 28 '13 at 21:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.