The only time we saw major performance issues with ephemeral (we're using SSDs in RAID 0) was when we ran fio against a sparse file. It sounds like you ran it against a properly filled file though, and it looks like you're on a single spinning drive, based on the fio numbers. Can you confirm? Also, what version of qemu are you running, and can you check if you have iothread enabled? "qemu-system-x86_64 -device virtio-blk-device,help" should show iothread somewhere. One other thing I can think of, did you specify directIO in your fio run?
I notice you have the balloon driver on. Have you had used it in a production environment that neared full? Curious what the feedback is here. Warren Warren On Wed, Jul 8, 2015 at 6:58 AM, Gleb Stepanov <[email protected]> wrote: > Hello, all. > > We have measured disk I/O performance on openstack virtual machines > with aid of > FIO tool. We've tested performance on root dist drive device, test > consists of write operationby 4kb > blocks to file with size 90Gb (prefilled in advance). > We use qcow2 image for vm, ephemeral drive and virtio driver. > All configuration goes in attachment. > > There are some results: > > test 1 > > threads 1, 5, 10, 15, 20, 40 > iops 72,58,49,60,94,72 > > test 2 > threads 1, 5, 10, 15, 20, 40 > iops 71,60,54,88,52,52 > > test 3 > threads 1, 5, 10, 15, 20, 40 > iops 71,49,58,51,128,130 > > test 4 > threads 1, 5, 10, 15, 20, 40 > iops 65,49,60,56,52,63 > > As it is shown performance degraded during increasing amount of > threads, also deviation of results on 40 threads is very big. > Have you any ideas how to explain performance behaviour? > > Kind regards, Gleb Stepanov. > > _______________________________________________ > OpenStack-operators mailing list > [email protected] > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators > >
_______________________________________________ OpenStack-operators mailing list [email protected] http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
