[zfs-discuss] Low read performance on ZFS based on iSCSI disk

kaazoo kaazoo at kernelpanik.net
Wed Jan 13 03:28:55 EST 2016


Am 12.01.16 um 18:43 schrieb Ruben Kelevra via zfs-discuss:
> Which io scheduler do you use? If you use cfq you should change it to none. :)
> 


Hello Ruben,

I guess you mean 'noop' instead of 'none'.

I switched from 'deadline' to 'noop' on both iSCSI paths and their
multipath device:

echo "noop" > /sys/block/sdd/queue/scheduler
echo "noop" > /sys/block/sde/queue/scheduler
echo "noop" > /sys/block/dm-2/queue/scheduler


However, that doesn't seem to fix the problem. The read performance with
ZFS is still low compared to directly reading from the iSCSI device.


Regards,
kaazoo


More information about the zfs-discuss mailing list