Re: [nmrg] Control plane [was Next version of draft-irtf-nmrg-autonomic-network-definitions]

[Brian E Carpenter <[e-mail]>]

On 30/07/2014 20:45, Michael Behringer (mbehring) wrote:
>> -----Original Message-----
>> From: Brian E Carpenter [mailto:[e-mail]]
>> Sent: 29 July 2014 22:22
>> To: Laurent Ciavaglia
>> Cc: Michael Behringer (mbehring); [e-mail]; [e-mail]; draft-
>> [e-mail]
>> Subject: Control plane [was Next version of draft-irtf-nmrg-autonomic-
>> network-definitions]
>>
>> Just commenting on one point from Laurent. I'll stick to cross-posting for
>> this point. In due course I think we should avoid it, but OK for now.
>>
>>> -It is not clear/straightforward what "in the global context of each device"
>> means. If we want to say that the control protocols can run in-band/out-
>> band, let's just say that.
>>
>> Yes, I think we have to be more explicit. One approach is to have an explicit
>> autonomic control plane which is an in-band overlay on the physical
>> network. Another approach is to make this implicit - just use the existing
>> L2/L3 network but without considering it to be a control plane at all (that's
>> what most routing protocols do). A third approach would be a truly separate
>> control plane such as a dedicated L2VPN. The point is that they could all
>> support autonomic behaviour; it's a design choice.
> 
> Suggestions for better wording is welcome. For a L3 device it's relatively clear: It's either the global routing table or a virtual routing table, such as a VRF, or a virtual router. Problem is that the concept of the Autonomic Control Plane as described in draft-behringer-autonomic-control-plane applies also to non-L3 devices, for example to switches or NMS systems, or .... 
> 
> So, what is the generic term for "global routing table" and "virtual routing table" which is valid also for non-L3 devices? 
> 
> So, the full range of options is: 
> - inband, as we run for example IGPs today
> - Over a configured VPN ("management VPN")
> - Over the self-managing Autonomic Control Plane (See above mentioned draft)
> - Over a "real" out of band network. 
> 
> Should we list those, 

IMHO, we should definitely list them.

> and explain what they do in a bit more detail? Maybe that would make things clearer... 

I'm not sure that is needed in the definitions draft; it might get a bit
complicated. But I certainly have no objection to one or two sentences
for each one. We should probably avoid pros and cons, since that is
not the job of a definition.

   Brian