Hello, Fo years I left the default setting of our HDs IO scheduler. After reading some articles about HW RAID and SAN it seems to me that it is not a good thing to let Linux doing IO scheduling because the HW between him and the HDs have his own IO scheduler.I will probably run iozone to test with default and NOOP for example but I am curious if someone has some toughts on that subject ?
Given all the ideas on the topic you find in the other posts we did some testing. Results are inconsistent:
- 5% advantage for noop with Dell Perc 4Di hardware raid controller, raid5 of 3 disks, RHEL4.
- 6% disadvantage for noop with RHEL5 running in a VMware ESX 4 based virtual machine. Of course IO-performance of ESX is crap anyways, so that may not really be a valid result. As it is a customers system, I do not remeber the storage make / model.
- More or less no difference with RHEL5 VMs and Xen Dom0 systems attached to multipathed FC SAN, Eonstore RAID enclosures (Raid6 of 8 SAS disks), Qlogic HBAs and FC switches.
However, there is more to think of when it comes to IO performance than the scheduler: various block sizes at different positions of the IO stack: - harddisk block size (used to be 512 byte, but 4k blocks are round the corner...)
- raid controller stripe size and RAID type / layout - maybe an FC / iSCSI protocol or HBA block size - LVM stripe size - filesystem block size - OS readahead settingsGuess one has to align all of that to achieve maximum performance, we did not extend our test that far yet...
best regards, Gunther -- Gunther Schlegel Manager IT Infrastructure ............................................................. Riege Software International GmbH Fon: +49 (2159) 9148 0 Mollsfeld 10 Fax: +49 (2159) 9148 11 40670 Meerbusch Web: www.riege.com Germany E-Mail: [email protected] --- --- Handelsregister: Managing Directors: Amtsgericht Neuss HRB-NR 4207 Christian Riege USt-ID-Nr.: DE120585842 Gabriele Riege Johannes Riege .............................................................YOU CARE FOR FREIGHT, WE CARE FOR YOU
<<attachment: schlegel.vcf>>
_______________________________________________ rhelv5-list mailing list [email protected] https://www.redhat.com/mailman/listinfo/rhelv5-list
