>""Haakon Claassen (hclaasse)"" wrote in message >[EMAIL PROTECTED]">news:[EMAIL PROTECTED]... >> Perhaps the Multi protocol >> >> Is in regards to the fact that it can support multiple routing contexts >> (one per vrf) > >That's a pretty weak definition of 'multiprotocol'. > >More to the point, even if you're talking about RFC2547 vpn's (which is only >a subset of MPLS functionality), you still require IP in the core. Why is >that required? Why can't I, for example, build RFC2547 vpn's on an ATM >core, where my ATM switches do not speak IP, but do speak a (theoretical) >version of MPLS that is completely compatible with ATM dynamic signalling?
That's almost exactly what Ipsilon did with IP switching. If for no other reason, they ran into scaling problems, because they needed a VPI/VCI field for every flow. > >Now you might say that I could do this by just installing IP edge (PE) >routers over an ATM core, and the PE routers peer to each other with IP and >MPLS, and the ATM switches peer to each other with PNNI. But that sucks. >The whole promise of MPLS was to offer a unified control-plane. Current architectural thinking is that control planes are necessarily multilayered. Routing protocols and label distribution protocols, to say nothing about refinements in traffic engineering and failover, operate at different conceptual levels. For that matter, there are medium-specific control protocols below MPLS. > Not to >mention I still have the N-squared scaling problem with my edge routers. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=54568&t=54507 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
