The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2001-Jan> msg00066



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Some queries

  • From: Alan Hannan <alan@routingloop.com>
  • Date: Sat, 6 Jan 2001 13:02:10 -0800


  It is useful to point out that within a domain
  such as the backbone, it is very common to have a partial
  mesh at L1 (circuit), and a full mesh at L2 (ATM PVC/sPVC or MPLS
  LSP) and L3 (IP Routing and/or forwarding adjacency)..

  Note however that all routers in the network/AS may not participate
  in a full mesh at L2/L3 (much les L1) due to hierarchy.  For example,
  Customer aggregations routers, even if they perform MPLS, are 
  unlikely to have a full mesh w/ all other routers in the AS.

  Additionally, it is not uncommon to find large/national/global 
  networks will have hierarchical domains, say, of Global Level, 
  National Level, Regional Level, and Metro Levels.  Within each 
  of these domains one can expect partial mesh at L1 and full mesh 
  at L2/L3.

  In such a network, you might find 40 routers in 20 pops on the 
  top-most backbone level, w/ each router having an average degree of 
  freedom of say 2.0 and each POP of 4.0.  Therefore you would expect 
  and observe about 80 physical backbone links (DOF*N/2) and about 
  39 L2/L3 adjacencies  (M-1).  Note that there should be (M*(M-1)) or 
  about 1560 total forwarding paths in the backbone domain alone.

  -alan


Thus spake Randy Bush (randy@psg.com)
 on or about Sat, Jan 06, 2001 at 12:06:35PM -0800:
> > Do you mean to say SPs such as large/national ISPs use ATMs on the backbone
> > and have complete mesh with DS3/OC-3/OC-12 links 
> 
> no, complete mesh of vcs over semi-meshed physical circuits
> 
> randy