On 09/07/2009 05:46 AM, Yao Wei wrote:
> The main purpose of my using iscsi is to implement remote disk
> copying.
> When I installed iscsi-target on Dell machine and using the whole disk
> to be the target:
>   Lun 0 Path=/dev/sda,Type=fileio
> then on my ibm x3550 machine (with two NICs) , I first configure  NIC1
> and NIC2 with different MAC addresses in /etc/iscsi/ifaces/iface0 and
> iface1, and then I  use iscsiadm -m discovery -t st -p ip -I iface1 -l
> to discovery and login the iscsi discovery.
> To test the  write and read speed of remote disk, I first use hdparm -
> t /dev/sdb to get the speed of timing buffered disk reads: less than
> 10MB/sec.
> More, I directly using dd if=/dev/sdb of=/home/my.dd ibs=32k to copy
> the whole disk, it consume very much time, as slow as 10 MB/s.
> Later, I let IBM x3550 machine to be the iscsi-target, and Dell
> machine to be the initiator, and do the procedure as above, I found
> that the speed of hdparm command can obtain about 80-100 MB/s, and
> also the command of dd a 40GB hard disk can be finished within
> 15 minutes (about 50 MB/s).
> I found that as long as I use other machine(only with one Nic), it
> will not emerge the above circumstance. That is to say, whatever it be
> an initiator or target, the remote read and write speed will not be
> quite different. But, on machine with two NICs, it will perfor so
> different.

I am just saying it might be the disk that is used here:
Lun 0 Path=/dev/sda,Type=fileio

If you set up the target on the x3550 and use
Lun 0 Path=/dev/sda,Type=fileio
then on the target box use hdparm to read /dev/sda what values do you 
get? Is it close to what you get when you read it from the initiator 
through iscsi remotely?

I was suggesting that to make sure the disk being used for the 
iscsi-target/IET disks is the same speed you can take the disk out by 
doing this:

Target iqn.2001-04.com.redhat:perftest
        Lun 0 Type=fileio,Type=nullio,Sectors=10485760

With this when you test both boxes at a more level playing field, and 
you can begin to narrow down what component is creating the throughut 

Did I understand you right?

You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To post to this group, send email to open-iscsi@googlegroups.com
To unsubscribe from this group, send email to 
For more options, visit this group at http://groups.google.com/group/open-iscsi

Reply via email to