[Nml-wg] Modelling EPL and its relationship to EVTS

[Jeroen van der Ham]

Hi,

I think this is an interesting question that is an excellent use-case for NML, so I'm forwarding it to that group also. And I would like to invite everyone into this discussion also.

On 10 Oct 2013, at 17:24, John MacAuley <john.macauley at surfnet.nl> wrote:

> Jeroen, Freek,
> 
> I am trying to come up with a way to describe an EPL service port (802.1ah UNI-N) and its relationship to an EVTS service port (802.1Q Trunk).  I have included the following NSI topology picture for illustration.
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PastedGraphic-1.pdf
Type: application/pdf
Size: 47533 bytes
Desc: not available
URL: <http://www.ogf.org/pipermail/nml-wg/attachments/20131016/05d62f32/attachment-0001.pdf>
-------------- next part --------------
> 
> <PastedGraphic-1.pdf>
> Just to make sure we are clear on terminology:
> 
> - The EVTS service is a switch Ethernet service based on the 802.1Q Trunk specification.  A single VLAN tag is used to identify related Ethernet flows and perform switching.
> - The EPL service is a Metro Ethernet service definitions taking all tagged and untagged traffic arriving at a UNI-N port and transparently passing to a paired UNI-N port using Carrier Ethernet (802.1ah) in the middle.  Once a port pair is "connected", all traffic traversing those ports is bridged.  No further connections can be made to different ports on these two ports until the existing connection is removed.  It is possible to filter only a subset of VLANs if full bridging is not desired (equivalent to a subset of 802.1Q STP).  Additional VLANs can be added to the filter is needed (equivalent to connecting new 802.1Q STP between the two paired ports).
> 
> So here is my problem…  How do I model this relationship in NML? 
> 
> 1. There is adaptation occurring at STP Y1, Y2, and Y3.  The NSI topology model is really not worried about that, but I think we should model it anyways.  How would I use adaptation and deadaptation to model this?

My solution would be to use regular Links (or isAlias in NSI) between the adjacent ports, and then define adaptations at Y1 and Y3 from 802.1Q to 802.1ah.
Because Y2 is not connected to anything it is hard to model an adaptation for it, so we'll just ignore it.

> 2. There are potentially 4K SDP on each EPL port, however, once one is used all others must be interconnected to STP on the same exiting port if use of that STP is desired.  This is a dynamic blocking switch matrix problem.  Can I model something like this in NML using the SwitchingService?

I don't think you necessarily have to model the 4k SDPs. The filtering can also be represented by the adaptation function using the labels.
I'm not sure whether you would necessarily need the SwitchingService in between, but it is a good thing to add anyway (and allows you to define other logic later on as well).

> 3. We need a way of tagging a port with a specific Ethernet capability so I can tell the port associated with STP X2 is an 802.1Q Trunk port and STP Y1 is an EPL port.

Agreed, we should define something for this either in the ethernet namespace, or in a separate namespace.

> 4. I assume I can use the isAlias relationship to interconnect STP X2 to STP Y1?

Sure.


Jeroen.