[Beowulf] raw ethernet

[Jeff Layton]

Robert G. Brown wrote:

>On Tue, 20 Jul 2004 050675 at student.unife.it wrote:
>
>  
>
>>Hi all, I'm making some studies (for my degree) about raw ethernet on a P4
>>pentium based cluster. Is someone interested in this? If yes I'll be happy
>>to share opinions and ideas (I'm not an expert, only a student BTW if you
>>want i'm here).
>>    
>>
>
>Do you mean using raw ethernet packets on a flat bridged network to do
>point to point IPCs in a cluster application?  No TCP/IP?
>
>Sounds like a lovely project.  Why don't you report to the list once a
>month or so.  Just remember, while you can likely skip the IP part
>(which is primarily associated with name resolution and routing) and can
>maybe skip the port part (dedicated function packets) you'll still have
>the rest of TCP to recreate -- reliable packet transmission and
>sequencing and checksumming.  Pretty ambitious, if you're planning to
>write the whole communications stack yourself (but yes, a great learning
>project).
>
>   rgb
>

I belive that HyperSCSI does this as well as Scali MPI Connect in ther
DET (or is it DAT) mode of running. Since HyperSCSI is open source
you might want to look at their code to see what they did.

Good Luck!

Jeff

>>Simone Saravalli
>>    
>>


-- 
Dr. Jeff Layton
Aerodynamics and CFD
Lockheed-Martin Aeronautical Company - Marietta