IOPS

[esxifreak]

hi a||

i have setup my personal lab. i am using openfiler as iscsi storage and when i tried to clone the vm it is taking long time to clone, what might be the issue.

My Open Filer PC:

CPU: 3.0Ghz

Ram: 1GB

HDD: 80GB SATA = 320GB IDE x 2 ( as mirror)

how to check what might be the issue.

Many Thanks in advance

KaDer

[rickardnobel]

esxifreak wrote:

HDD: 80GB SATA = 320GB IDE x 2 ( as mirror)

how to check what might be the issue.

Could you clarify some more on the hard disk setup? The numbers does not really seem to end up.

What kind of performance do you get when cloning the VM?

You could use tools like IOmeter to test the overall iSCSI performance.

[esxifreak]

Hi Richard

Thanks for the reply.

I setup the pc as below.

HDD: 80GB SATA (Open Filer Installation)

HDD: 320GB IDE x 2 (Configured as Software Raid1)

I connected the openfiler as iscsi in the host and when i try to clone the vm which i created first in the host took around 45mins to clone. Is it a normal speed or something i have to do.

Thanks

KaDer

[rickardnobel]

esxifreak wrote:

when i try to clone the vm which i created first in the host took around 45mins to clone. Is it a normal speed or something i have to do.

How large is the VM you are cloning? It will depend on that naturally, but 45 minutes seems like a quite long time. How have you setup the R/W Mode and the Transfer Mode in Openfiler? That will affect the write performance a lot. See this post for some details.

[EdWilts]

SATA drives are always going to be disappointing for performance-based operations like cloning.

Each spindle is capable of roughly 100 IOPS assuming modern 7200 RPM drives.  Given that these are IDE drives, and in a software RAID configuration at that, that could be stretching it.

SATA is also uni-directional - you can either read or write but can't do both at the same time.  That means that each read opeartion will have to complete before you can start the write operation and they won't overlap (unlike SCSI).  If you're doing a clone to the same destination as the source, this is going to be slow.

[rickardnobel]

Ed Wilts wrote:

Each spindle is capable of roughly 100 IOPS assuming modern 7200 RPM drives.

That would be for random IOs, for cloning it might be somewhat better since it mostly involves sequential read and writes, however of course two SATA drives will have a certain and quite low limit of throughput.

Happy New Year! Off to party, see you next year.