RE: Architectural scope of IP over Inifiniband

>It would greatly help if i can get hold of the first BOF.
>Is there a link i can follow.

See IETF meeting archive at
http://www.ietf.org/proceedings/01mar/index.html

IPoIB gives you the complete "IP semantics" (both from the API and
network management perspectives) and it uses UD (unreliable datagram)
as the default transport mechanism at the IB layer like Bill said.

However the IPoIB WG is also chartered to support IPoIB over IB
connections (see the WG web page). You may view this as tunneling
IP packets across an IB fabric between nodes sharing an IB
connection.

Jerry

> 
>Thanks, John
> 
>-----Original Message-----
>From: [email protected] [mailto:[email protected]]On Behalf
>Of Michael Krause
>Sent: Monday, September 15, 2003 7:03 AM
>To: [email protected]
>Subject: Re: [Ipoverib] Architectural scope of IP over Inifiniband
> 
> 
>At 03:09 PM 9/13/2003 -0700, [email protected] wrote:
> 
> 
>>I was actively following the RDDP group, and happened
>>to come across this interesting group as well.
>>
>>I am kind of lost as to why one would want IP which is
>>unreliable to be carried over a reliable link
>>infrastructure without any reliable transport layer at
>>the TCA or the HCA.
>>
>>Would the intent of this be to tunnel the IP packets
>>through a Infiniband fabric and emerge as IP to some
>>point outside that supports TCP/IP for reliable
>>transport.
> 
>The intent is to allow the IP protocol suite to be supported over an
>InfiniBand fabric - there are many usage models even with an InfiniBand
>only solution (this was covered in the first BOF to kick off this workgroup
>- not sure where those slides are posted but might be a good background
>reference). This can be between two endnodes, e.g. a set of servers, or
>between two devices, e.g. an IB-to-IP gateway where one links an InfiniBand
>connected set of endnodes with another fabric type (e.g. Ethernet) and
>provide IP routed packets between the two domains.
> 
>Mike
>