ib-iser - 4k IO req size only

ib-iser - 4k IO req size only

[Heiner Lesaar]

Dear all,

I use SCST 3.3.0  and ISCSI-SCST on CentOS7.3 (stock kernel) w. in-tree RDMA  and Mellanox ConnectX-4 cards (no OFED packages) on the target.

Same hardware / software on the initiator (w. open-iscsi and ib-iser, no OFED as well).

When using iSer as the transport layer, all read and write requests from the connected initiator to the target devices are always exactly 4kb (checked w. e.g. "iostat -xm 2"), which harms the performance of large IO a lot obviously. 

Changing to TCP as the transport layer shows expected behaviour of larger IO requests (up to max_sector_size) and expected performance.

Tested w. common everyday operations like "cp" or "dd", but also benchmark tools like "filebench".
I guess its something related to ib_iser, so I hope someone on the list has a good hint for me how to solve this.


Thank you very much for any kind help or response!


Heiner

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: ib-iser - 4k IO req size only

[Max Gurtovoy]

hi Heiner,

On 4/8/2017 12:59 PM, Heiner Lesaar wrote:
> Dear all,
>
> I use SCST 3.3.0  and ISCSI-SCST on CentOS7.3 (stock kernel) w. in-tree RDMA  and Mellanox ConnectX-4 cards (no OFED packages) on the target.
>
> Same hardware / software on the initiator (w. open-iscsi and ib-iser, no OFED as well).
>
> When using iSer as the transport layer, all read and write requests from the connected initiator to the target devices are always exactly 4kb (checked w. e.g. "iostat -xm 2"), which harms the performance of large IO a lot obviously.
>
> Changing to TCP as the transport layer shows expected behaviour of larger IO requests (up to max_sector_size) and expected performance.
>
> Tested w. common everyday operations like "cp" or "dd", but also benchmark tools like "filebench".
> I guess its something related to ib_iser, so I hope someone on the list has a good hint for me how to solve this.

I sent a fix in 1e5db6c31ade41 commit.
please apply it.

Max.

>
>
> Thank you very much for any kind help or response!
>
>
> Heiner
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html