I don't fully understand... If you have a pair of disk (two) you can't
make RAID 0+1 (you'll need at least four for this raid level). If you're
using SATA disks (75-100 IOPS per disk depending of vendor and quality),
if you're reading from both you can get 200 IOPS maximum, 100% reading of
course, if you mix read and writes, if you have mirrored disks, when you
make a disk write IOP, this will be translated in 2 IOPS (one on each
disk). Depending of the RW percentage, IOPS can vary.
If you have four SATA disk (two pairs), your higher speed is 400 (100%
reads). So, if you mix writes, in this case, each writes will generate 2
IOPS at disk level... Again, depending of the Read-Write ratio, your
result can be consistent.
Looks like that you can have IO limitations... The solution could be have
more disk in the array group to increase the available bandwidth.
On 10/11/10 12:36 PM, "Communities emailer"
Yes I know we have an I/O limitation, that's why I'm posting here.
Sorry for the confusion HP says in the tools that it's RAID 0+1 but really it's just RAID 1 (mirroring), I really wish HP would be clear about this.
So we actually have 2x 300GB Serial Attached SCSI (SAS) 6g Disks configured as RAID 1.
I want to know if this kind of throughput is normal for this configuration, and if not how I can find the bottleneck.
Two sas disk can offer a maximum of 400 iops more or less, depending of vendor. So, if you make, i.e., a 50% read/write workload with 150 iops generated by vm, in the disk you'll have 225 iops in disk (75 read, 150 write). Depending of the read write ratio, if you have more writes, you can have io problems. These are rough numbers...
Sent from my iPhone