On Thu, 18 Jun 2009 12:12:49 +0200
Hannes Reinecke <h...@suse.de> wrote:

> 
> Hi all,
> 
> Joachim Worringen wrote:
> > On Jun 18, 11:19 am, Boaz Harrosh <bharr...@panasas.com> wrote:
> >> On 06/18/2009 10:56 AM, Joachim Worringen wrote:
> >>
> >>> Greetings,
> >>> I tried to use Open-iSCSI with a non-tcp socket type and failed
> >>> (timeout after connection has been established).
> >>> Looking at the source, the reason is obvious: for sending data
> >>> (iscsi_send()), the function pointers from sock->sk are used via the
> >>> kernel socket API. This works well with non-tcp sockets. Howver, for
> >>> reading data (see callback handler iscsi_tcp_data_ready()),
> >>> tcp_read_sock() is used instead of the related kernel socket API call.
> >> Sounds good. Could you test out your solution and send a patch?
> >> If it tests out, I don't see why not.
> > 
> > We don't have a solution yet, just a problem...
> > 
> >>> Is there a specific reason for this? iSCSI would surely benefit from
> >>> using high-performance, non-tcp sockets if available (I'm talking
> >>> about SuperSockets in this case, 
> >>> seehttp://www.dolphinics.com/products/dolphin-supersockets.html).
> >> Sure sounds nice. If it is a simple and compatible change it sounds very 
> >> good.
> > 
> > I haven't looked into it enough to claim it'll be a simple change. I
> > figure the iscsi recv code will change completely, but would be
> > simplified as we don't need to deal with the tcp buffer details etc.
> > when using a simple socket recv call, but it remains to be seen how
> > this affects compatibility, i.e. conc. partial or interrupted
> > transfer.
> > 
> > It is not decided whether we allocate resources for this.
> > 
> Well, the obvious solution here is to implement another transport module (eg
> iscsi_supersockets.c) much like it's done for iSER.
> The rest of the code should be sufficiently abstracted to handle it properly.
> 
> There is a reason why it's called iscsi_tcp.c ...

Yeah, I guess that all niche high performance interconnect technology
(such as Dolphin, Myrinet, etc) for HPC support RDMA. They could do
something like iSER. They support upport kinda socket interface but
socket interface is not optimal for them. So modifying iscsi_tcp for
them in a strange way doesn't make sense much.

Well, anyway seems that 10GbE is slowly killing all niche high
performace interconnect technology.

--~--~---------~--~----~------------~-------~--~----~
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 
open-iscsi+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/open-iscsi
-~----------~----~----~----~------~----~------~--~---

Reply via email to