More 2c, for everyone. I had unacceptable speeds in general when I first started using open-iscsi, mainly manifested when reading from my IET target (both Ubuntu BTW). One issue was as another user mentioned, I was reading from a fast target (2x3Ghz Xeons) and writing to a slower (1x2.4Ghz P4) more taxed system running software raid. After I realized that the speed issue made more sense as the bottleneck was the writing to the software raid. Anyhow... I enabled jumbo frames on the NICs (MTU > 1504), specifically an MTU of 7000 as that is all the client NIC could handle. Although 9000 would've been best. This produced consistent reads from the target at 800Mbps-1Gbps (I was surprised too, used iperf for tests), and write speeds of around 500-650Mbps.

Again 2c.

Thanks for all the work devs!

D.A. wrote:

On Sep 12, 1:29 am, "A. Eijkhoudt" <[EMAIL PROTECTED]> wrote:
Hello and thanks in advance for reading,

We've been having serious read performance issues with Open-iSCSI.

I am having a similar problem, with good write performance and
dreadful read performance, but with bursts of good speed.

I am not entirely sure that this will be the final answer (too many
dead ends already), but for me it seems it was a mismatched block size
between the target and initiator devices (and upgrading the iSCSI
target). The target device had a block size of 4096, while on the
initiator side the block size was 512. This issue seemed to manifest
itself particularly with multipathing.

In my case I am using two Ubuntu boxes (open-iscsi v2.0-865 and IET
v0.4.15/0.4.16). I am using software raid (this is a test box) with
LVM, and the LV is presented to the iSCSI target.

Just my 2c.

You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To post to this group, send email to
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to