[openib-general] RE: [iSER]question about iSER code

[Alex Nezhinsky]

xg wang wrote:
>>     I read the draft-ieft-ips-iser-04 and the iSER source code. In
>> the draft, it wrote ...
>>    So I think that the iSCSI connection is different from the iSER
>> conection since it 'allocate the connection resources'.
>> Does it mean iSER layer will estabilsh another iSER-connection. I am
>> not sure. ...

Dan Bar Dov wrote:
> The ISER spec is iWARP oriented. Please read also the
> Huffered spec regarding ISER over IB. There is work in
> progress regarding ISER spec modification so it becomes
> transport neutral. The ISER implementation at openIB starts
> the connection Already in ISER mode, so the flow you list
> above is not followed. 

There is also a new draft candidate by Mike Ko, incorporating the 
changes proposed in the Hufferd's draft. It can be found as
draft-ietf-ips-iser-05-candidate.pdf or 
draft-ietf-ips-iser-05-candidate.txt 
at http://www.haifa.il.ibm.com/satran/ips/

This text refers to the transport layer as an RDMA Capable Protocol,
mentioning the differences between iWARP and IB when necessary.

iSER layer does not need to establish a new connection. ISER serves as
a transport layer for iSCSI, so its connection is *the* connection, at
least 
for the etire full-featured phase. The only issue is related to the
login phase 
but this is being discussed here and in IETF-IPS.

Alexander Nezhinsky
 
Software Engineer
Infiniband Storage Solutions
<http://www.voltaire.com/> 
9 Ha-Menofim, Herzelya 46725 Israel
tel: +972- 9- 9717637
fax: +972- 9- 9717660
mobile: +972-50-7504376