On Tue, Nov 14, 2000 at 07:30:57PM +0530, Vinod Anthony Joseph Cherunni wrote:
> Hi All,
>
> I am really sorry for taking up much of all your time, Just a few queries
> that I have, If possible pls take a little time for my queries.
>
> When RSVP is used to achieve traffic engineering, Typically I understand
> that it would request the network for a configured amount of bandwidth
> when a conversation starts. Now If I configure a VPN to be allocated 2
> Mbps bandwidth, When a conversation starts will RSVP request for 2 Mbps
> bandwidth & not release excess bandwidth that is not used during the
> conversation, or else does it work statistically, wherein wasted bandwidth
> is released during the conversation.
I'm not sure, but I think you may have VPN and TE confused. They are
two different things, although you can certainly apply TE to a VPN
(but it's a bit more work). But yes, for an MPLS-TE LSP, 2MB (or
whatever) is requested and held. MPLS-TE is designed to make
aggregate reservations, not per-flow, so there's no concept of
dynamic call setup/teardown. This is one reason it's not always a
good idea to make per-VPN reservations; you don't really want to end
up with hundreds of tiny reservations across the backbone. That's not
what MPLS-TE is for.
> "IIRC It doesn't work statistically and has options for pools that
> you can configure and you can add different "classes" for which pool
> the bandwidth is allocated from based upon what COS you want to
> offer." - Can this be explained in a little more detail, I would
> greatly appreciate if it be explained with an example pls.
This feature isn't out yet, but will probably be out sometime soon.
When it comes out, dig up the documentation on www.cisco.com, and
that'll have examples in it.
> I understand that people offer MPLS VPN services using Frame Relay
> as the last mile to the Customer. This service provider builds a
> Frame Relay access & an MPLS core typically using a switch like the
> Cisco BPX. The benefit here is that the customer is only using a
> single PVC, & can get connected to how many ever Intranet sites
> needed.My question is in such a scenario how does the service
> provider offer lastmile CIR's (comitted bandwidth) for connectivity
> to multiple sites using a single PVC. In the acase of IP probably CAR
> would do the job at the ingress.
One of the two of us is extremely confused. :)
What you use as your last-mile connect has nothing to do with whether
or not you use MPLS in the backbone. If you connect a customer to a
FR switch, and you're not actually speaking MPLS-TE with that customer
(which you shouldn't be, in general), you use the same CIR
provisioning methods you always have with FR.
Or are you asking something different?
eric
>
> Kindly advice,
>
> With regards,
> Vinod.
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:21 EDT