RE: DS-TE overbooking model - Not that complicated
"Yuji KAMITE" <[email protected]>
| Newsgroups | gmane.ietf.tewg |
|---|---|
| Message-ID | <[email protected]> |
> > But to guarantee minimum bw for an LSP or for a CT as its LSPs are set > up over time, there's got to be a way for upstream nodes to learn the > LSP under-sizing factor that is in use. > > I'm hoping that instead of LOMs and instead of LSP under-sizing, BC > over-sizing will be deemed sufficient to support "per-CT overbooking"! I think LSP under-sizing is difficult solution, too. LSP under-sizing approach will be burden on operators because it will occur significant discrepancy between LSP's size and expected traffic size. (In other words, for practical usage, complicated implementation must be done such that it can take care of all these under-sizing factors at upstream and can adjust rousources if necessary, automatically and dynamically.) In my opinion, it will be easy to start with LSP size is set the same as expected bw ( this value is not under-sized). if overbooking is necessary, BC is adjusted first. seems quite simple. From the same reason, I think the place where LOM is used might be very small at first deployment. But, I'm not sure if closing LOM itself is really valid answer now. Perhaps it is useful for changing reservation capacity temorarily for a short time For example, due to link construction or rapid load increase, SP might want to change capacity temporarily althogh it does not want to change BC or LSP size itself. ( because of SP's policy.) Actually SP would select the best solution from various perspective. While avoiding complexity might be a key factor for some SP, another SP will flexibility is important. However, as for LOM, I'm wondering where it can be applied and useful. Actually, I'm not sure the example written above is good one. Today there seem few applicability suggenstions. Requirement says DS-TE must support overbooking, but is LOM regareded necessary and sufficient ? Possibly other approaches can provide the same function. In order to support LOM, LOM's advantage and applicability must be clarified. Any ideas? --- Yuji KAMITE (e-mail: [email protected]) Innovative IP Architecture Center, NTT Communications Tel:+81-3-6800-3261, Fax:+81-3-5365-2990