At 3:29 AM -0500 3/22/02, Tom Scott wrote: >Kent, Irwin, > >This is an interesting exchange of perspectives. Could you >or someone else comment on GMPLS?
GMPLS does many of the things you are describing. It "generalizes" MPLS setup beyond the current packet/frame oriented paths to paths that don't have a granularity as fine as packets: optical wavelengths, multiplex time slots, and specific physical port sequences on various devices. I really don't see MPLS/GMPLS as a relevant technology for most enterprises. Let me throw out an analogy that I just invented -- I just got up and may not be thinking clearly. 1. ATM was originally intended as a carrier-only technology. FR, ISDN, etc., were seen as customer access protocols to the carrier ATM cloud. ATM UNI specifications came into being due to a market need for a high-speed access technology. 2. MPLS often is called "ATM without cells." I consider (G)MPLS analogous to the original carrier role of ATM, and things such as provider-provisioned VPNs (PPVPN) at both layer 2 and 3, metro optical Ethernet, 3G wireless, etc., corresponding to the access protocols in the original ATM model. 3. MPLS is not a panacea, but does have many useful features for traffic engineering and fault tolerance, especially when dealing with very large numbers of L2 emulated circuits and L3 private networks/Internet access/large provider cores. GMPLS provides a smooth path for integrating both present (e.g., POS, SONET) and evolving optical technologies. How many enterprises will have a requirement to manage many fibers containing many DWDM wavelengths at OC-192 or OC-768? 4. There are enterprise needs that are just starting to get integrated with MPLS, such as IPsec. >How does it factor into >the comparison of MPLS vs. FR? Is there anything about the >combination of MPLS / GMPLS that gives it an advantage over >FR? Some market research I've seen suggests the telcos do not expect to have the IP-literate staff to do more than deliver the core and perhaps 10% of their VPNs as L3. L2 VPNs (e.g., frame and virtual wire emulation) is attractive to them because it significantly reduces their support costs. > >Another issue I'd like to understand in this context is >native MPLS transport. Do you see a possibility in the >future for simplifying the transport of MPLS packets? In >other words, could one replace SONET/SDH with a simple >transmission layer X (whatever that might be)? The stack >,might look something like this: > > 7 7 > 6 6 > 5 5 > 4 4 > 3 3 3 3 > 2 2 2 2 2 2 > 1 1 1 X 1 1 1 > > ^ ^ > | | > +-----+ > native > MPLS > core That's GMPLS, which specifically is intended to be able to transport SONET, POS, etc. > >There in the middle, where the "native MPLS" core would be >in a greenfield network, is it possible to transport the >MPLS packets on a fiber medium, possibly on different >lambdas using GMPLS, but without SONET/SDH? I don't work on >that layer (physical L1) enough to know the interactions >between L1/L2, but it seems that simplification is >desirable. Possibly using MPX (MPLS over PPP over X) instead >of traditional POS? > >There are many reasons to keep SONET/SDH, for example, >protection switching. Is it possible that MPLS / GMPLS could >offer similar solutions that would have a competitive >advantage? Maybe someone at MPLScon will have an answer. See >you there. Yes, and probably more advanced protection switching with more efficient resource use. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=39155&t=36670 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]