[j-nsp] MPLS in the Access

[Pavel Lunin]

> I think the biggest problem with this approach, is the limit of 24K
> MAC addresses across the entire VC. This is a problem we'll be running
> into in the not so distant future - I don't know if CCCs are the right
> (only?) solution to this.
>    
Right. This scheme has quite appreciable scaling limits.

If the next tier is DSLAM/MSAN for residential broadband, the avg of 100 
MACs per port (even a bit higher since you most probably not use all 24 
ports on the two 'centeral' boxes) seems normal. If you connect business 
customers right to the VC ring, this is not that bad as well, though not 
an infinity. But if you want the next tier of access switches where you 
aggregate the customers' links (say another x24), well, this can be a 
problem: 24k/240/24 = just 4 MACs per port or 8 if you use LAG to 
connect those switches to different VC-members. So if you want to 
provide multipoint L2 services for business customers, most probably 
this is not enough in such a 2-tier case.

CCC is a solution but VC losses half of its benefits, if you add the 
second tier here. Anyway adding another tier of cheap switches here 
seems to be a way to get the old good switching garbage.