On Sat, Nov 04, 2000 at 12:50:05PM +0530, Vinod Anthony Joseph Cherunni wrote:
> Hi All,
>
> Just another query on MPLS Traffic engineering.
>
> In order to ensure bandwidth guarantees for a VPN customer, I understand
> that RSVP signalling could be used. The cisco web site states that for
> achieving the same, tunnels need to created & RSVP signalling used, The
> other point was that IS-IS is the only protocol that has the necessary
> extensions for MPLS. Now for a network like ours which is built purely on
> OSPF as the IGP, How will I achieve the same?.
MPLS TE with OSPF is supported as of 12.0(9)S or 12.0(10)S.
> Similarily is'nt this
> method of manually creating & maintaining tunnels an overhead, It
> resembles the previous generation IP VPN services using IPSEC encryption,
> wherein tunnels had to created & maintained seperately, resulting in
> management overhead. It will be great if I could get across some
> references for deployment.
>
> Another concern is how I limit bandwidth on the last mile of a customer.
> Given an example wherein a customer has the following requirement - He
> needs one connections to two of his branch offices & one conection to the
> Internet with the follwing bandwidth requirements -
> - Connection to branch A @ 1 Mbps (India to UK)
> - Connection to the Internet @ 1 Mbps (India to Internet)
>
> Assuming my network is built with two data centres, One in India & the
> other in UK, wherein the data centre in UK peers into AT&T for the
> Internet part, & the customer requires Intranet connectivity between India
> & the UK i.e. between his offices.
>
> Now If I am to provide him with a single last mile connection using an E1
> (2.048 Mbps) circuit into my PE router on which all his connections
> terminate. I would also need to control how is bandwidth is allocated
> end-end. Now internal to my Service provider network I will use "traffic
> engineering" for guaranteeing bandwidth. But on the Ingress (ie on the
> last mile) How will I ensure that Internet traffic is given 1 Mbps
> bandwidth, & the other circuit is allocated 1 Mbps. Just to ensure that
> one connection does not deprive the other connection of its bandwidth.
> Will I need to use something like CAR at the Ingress, or can MPLS traffic
> engineering be extended to the last mile.
You'll need to use CAR on the ingress to police this. 12.1(5)T will
see the release of Dual Bandwidth Pools, which may or may no tmake
your life a little easier.
>
> I understand that the VRF table decides the list of destinations that are
> allowed to be accessed by a particular VPN customer's network. Now if the
> same customer needs Internet access, How will it provisioned, will it be
> achieved by a default route at the customer end, which I feel will not be
> ideal. How will the VRF need to be populated?
>
You can either
a) put Internet routes in a VRF and import those routes into your user
VRF (not scalabale, bad idea, don't do it), or
b) 'ip route vrf foo 0.0.0.0 0.0.0.0 x.x.x.x global' where x.x.x.x is
a route known in your global table; that will get your VPN packets up
to the global routing table.
eric
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:20 EDT