[c-nsp] Seamless MPLS interacting with flat LDP domains

[Radu-Adrian FEURDEAN]

On Fri, May 3, 2019, at 04:15, Robert Raszuk wrote:
> Radu,
> 
> The MPLS in modern DC is none starter purely from technology pov. 
> 
> In modern DCs compute nodes are your tenant PEs all talking to rest of 
> the fabric L3. So if you want to roll MPLS you would need to do that to 
> the compute nodes. That means that with exact match you will see in 
> MSDCs millions of FECs and millions of underlay routes which you can 
> not summarize. Plus on top of that an overlay say L3VPNs for the 
> tenant/pods reachability.

Hi,

That is a specific design for a specific DC size. And even so, I see a problem with the compute nodes being the PE rather than CE.
But you have a point for that case - MPLS is not for networks that grow to a certain high amount of routers (and in your case endpoints become routers).

> IPv6 but all line up to a single buffer. 

Talking about IPv6, do you see many DCs deploying IPv6 in a menaingful way ? Just my curiosity...

Those being said, I'm curious how many datacenter do you see with millions of hosts AND routing down to host level.
For what I have visibility to, DCs go up to thousands of hosts (tens of thousands is already big, hundreds of thousands is huge), and network usually stops on a network device in front of them. You may not call them a "modern DC", but that's the most common occurrence.  And even for the case of big and huge DCs, there is a question to be posed about where you pout a boundary  for the aggregation level ? Always at DC border ? Really ? At that size ?

-- 
R.-A. Feurdean