NTAC Peering Working Group

[Paul Schopis <>]

Ok the Reader's Digest menu response;

1. Destination not reached - Semantics perhaps, does a back hole count as a 
path? But point is understood.
2. DC use case - Agreed; having networks move flexibly with migrating VMs, 
real good case.
3. Dead Link - Not sure how having a control plane not in the box 
understanding not to use a link for the data plane is necessarily superior. 
It sort of depends, with crank back their are scenarios where it could, and 
others where it won't, i.e. a graph problem is still a graph problem. MPLS 
does that as well. Ironically it inherited that from ATM.  
4. Substrate comment by Michael - Agreed, it is not very interesting if all 
it is used for is IP transport. But not sure I get the comment about 
test/diagnostic packets, again I think depends on what it is your trying to 
diagnose. Could you elaborate?

Finale thought for today: For me the value of SDN isn't necessarily 
abstracting away the control plane, it is the ability to 
orchestrate/provision. For technology to succeed, it need to do something 
better, faster or cheaper, actually probably two of those. 

Well enough philosophy for one day.

Grover, what was it you where seeking in your original question?

Thanks,
Paul
 
Paul Schopis
Chief Technology Officer
OARnet
1224 Kinnear Rd
Columbus, OH 43212
Ph. 614-292-1956
email: 


________________________________________
From: 

 
[]
 on behalf of Michael H Lambert 
[]
Sent: Thursday, May 30, 2013 11:25 AM
To: 

 Routing WG
Subject: Re: [2013-l3-architecture] FOR DISCUSSION: IP over AL2S

On 30 May 2013, at 11:09, Steven Wallace wrote:

> Not sure I agree. I think it's more likely that the packet takes the 
> intended path or it doesn't get to its destination. One can argue that not 
> getting to the destination is taking the unexpected path :-)
>
> SDN means moving away from things like shared fate (routing info is 
> in-band, so if the link fails I know not to use link) and network nodes 
> deciding how to forward a packet based on their view of the topology. This 
> may not be the best thing in a  WAN for certain applications.

And in a fully-evolved SDN network (whatever that means), can we guarantee 
that diagnostic packets follow the same path as data flows (or even that the 
data flows are packetized in the traditional sense)?  If we view the 
SDN-based network as just another layer-2.5 (or -2) substrate to run layer-3 
on top of, I don't think the problem is very interesting.

Michael