[ofa-general] OpenSM Up-Down algorithm

[Hal Rosenstock]

Hi Sagi,

On Sun, 2007-06-10 at 11:59, Sagi Schlanger wrote:
> Hi,
>  
> I'm looking for some answers on Up-Down routing at OpenSM .
>  
> Is anybody familiar with a utility/procedure to find credit loops
> given a topology and routing settings?

I know there was at least talk of ibdiagnet (in ibutils) checking this.
Not sure if it is implemented (yet) or if it is routing algorithm
independent. Eitan ?

> Is there a handy spec describing the OpenSM Up-Down algorithm?

The OpenSM up/down routing is based on the following paper:

"Effective Strategy to Compute Forwarding Tables for InfiniBand Networks"
Jose Carlos Sancho, Universidad Politécnica de Valencia
Antonio Robles, Universidad Politécnica de Valencia
Jose Duato, Universidad Politécnica de Valencia

http://csdl2.computer.org/persagen/DLAbsToc.jsp?resourcePath=/dl/proceedings/&toc=comp/proceedings/icpp/2001/1257/00/1257toc.xml&DOI=10.1109/ICPP.2001.952046

> What is the scheme through which roots are defined on clos and non
> clos/fat tree topologies?

The admin can supply the roots via -a  <guid_list_file> option when
invoking OpenSM.

       Auto-detect root nodes - based on the CA hop length from any switch
       in  the  subnet,  a statistical histogram is built for each switch (hop
       num vs number of occurrences). If the  histogram  reflects  a  specific
       column  (higher than others) for a certain node, then it is marked as a
       root node. Since the algorithm is statistical, it may not find any root
       nodes.  The  list  of the root nodes found by this auto-detect stage is
       used by the ranking process stage.

           Note 1: The user can override the node list manually.
           Note 2: If this stage cannot find any root nodes, and the user did
                   not specify a guid list file, OpenSM defaults back to the
                   Min Hop routing algorithm.

> Is this algorithm always credit loop free?

It's supposed to be.

> How efficient is using this algorithm on non clos/fat tree topologies?

What do you mean by efficiency ? Also, are you asking about pure fat
tree or non pure fat tree (or both) ?

-- Hal

> Thanks for your cooperation,
> Sagi
> 
> ____________________________________________________________
> Sagi Schlanger  | +972-9-9717651 (o)   |   +972-52-2385154 (m)
> Software Engineer, IB Switch
> Voltaire – The Grid Backbone
>  
>  www.voltaire.com
> 
>  
> 
> ______________________________________________________________________
> 
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
> 
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general