I was a little skeptic about this hp faulty driver, so i made some test in my lab made of two identical MicroServer Gen8, same G2020T cpu, 16 MB memory and same raid-zero lun as datastore built on two WD black 1TB drives.
As testing VMs I used two clones of CentOS 7 minimal with fio 2.1.10 installed, testing to /dev/sdb1 (thick eager zeroed vmdk).
No other vms were running during test.
fio command lines used
fio --filename=/dev/sdb1 --direct=1 --rw=randrw --refill_buffers --norandommap --randrepeat=0 --ioengine=libaio --bs=8k --rwmixread=70 --iodepth=16 --numjobs=16 --runtime=120 --group_reporting --name=8k7030test
fio --filename=/dev/sdb1 --direct=1 --rw=read --refill_buffers --norandommap --randrepeat=0 --ioengine=libaio --bs=4k --iodepth=16 --numjobs=16 --runtime=120 --group_reporting --name=4kseqread
| host1 : scsi-hpvsa 5.5.0.100-1OEM | host2 : scsi-hpvsa 5.5.0-88OEM |
8k7030test | read : io=358192KB, bw=2971.9 KB/s, iops=371, runt=120529msec
write: io=153376KB, bw=1272.6 KB/s, iops=159, runt=120529msec | read : io=341192KB, bw=2832.1 KB/s, iops=354, runt=120438msec
write: io=145792KB, bw=1210.6 KB/s, iops=151, runt=120438msec |
4kseqread | read : io=5732.3MB, bw= 48'906 KB/s, iops=12226, runt=120024msec | read : io=43512MB, bw= 371'288 KB/s, iops=92821, runt=120005msec |
The randow rw test are equivalent and i even got some more iops with driver version 100,
but the sequential read test was very bad.
After downgrading hpvsa to version 88 and rebooting, 4kseqread test now gives me : read io=39374MB, bw=335981KB/s, iops=83995, runt=120005msec.
I think I'll stick with driver version 88.
