[c-nsp] MPLS - MP-BPG with multiple OSPF areas

[David Prall]

Livio,
As Adam stated, I was thinking of prefix-suppression in the context of using
a single Area. Or perhaps minimizing the number of ABR's.

If you still want to break your network up into multiple areas. You might
want to deploy P's that are the ABR, while all PE's are in areas other than
0. Therefore all traffic between PE's will be inter-area via area 0 or
intra-area. So no traffic is directly destined to Area 0.

David

--
http://dcp.dcptech.com


> -----Original Message-----
> From: Livio Zanol Puppim [mailto:livio.zanol.puppim at gmail.com]
> Sent: Wednesday, October 19, 2011 5:53 AM
> To: David Prall
> Cc: cisco-nsp at puck.nether.net
> Subject: Re: [c-nsp] MPLS - MP-BPG with multiple OSPF areas
> 
> You're right David. There are out of order no packets, only
> asynchronous traffic. Sorry about that...
> 
> I don't think that only the supression would do the job, since the
> loopbacks will still be in different areas.
> 
> 
> 2011/10/18 David Prall <dcp at dcptech.com>
> 
> 
> 	Livio,
> 	Where are you getting out of order packets? You do have
> asymmetric hop
> 	counts, which most likely means asymmetric latency. But all the
> packets
> 	should be in order. Could use DWDM so that each router isn't
> directly
> 	connected and everything looks the same number of hops away, of
> course more
> 	ports are required at the Area 10 edge.
> 
> 	You can use prefix-suppression to only advertise the loopback
> using OSPF, to
> 	minimize the number of LSA's. Then use MPLS and BGP for all
> packet
> 	forwarding, including the global table.
> 
> 	David
> 
> 	--
> 	http://dcp.dcptech.com
> 
> 
> 	> -----Original Message-----
> 	> From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-
> 	> bounces at puck.nether.net] On Behalf Of Livio Zanol Puppim
> 	> Sent: Tuesday, October 18, 2011 9:34 PM
> 	> To: cisco-nsp at puck.nether.net
> 	> Subject: [c-nsp] MPLS - MP-BPG with multiple OSPF areas
> 
> 	>
> 	> Hello everybody,
> 	>
> 	> I have a doubt with a lab design that we are creating to test
> some MPLS
> 	> topologies. I would like to know if anybody can help me solve a
> problem
> 	> that
> 	> I am facing about routing paths. To help ilustrate the topology
> I'm
> 	> sending
> 	> the image link below.
> 	>
> 	>
> https://docs.google.com/leaf?id=0B4Hf34G524HsNTA3ZTc1NTItNmJlNi00ZDQyLW
> 	> I1ZDAtYTg5MTliODRjMDhk&hl=en_US
> 	>
> 	> In the topology, I have two core routers interconnected with a
> 1 Gbps
> 	> link
> 	> and several other routers interconnected with a 100Mbps link.
> The
> 	> interfaces
> 	> between the core routers are in the OSPF area 0 and all other
> physical
> 	> interfaces are in the OSPF area 10. Each one of the two core
> routers
> 	> also
> 	> have a connection to the area 10 using a 100Mbps interface.
> 	>
> 	> The router ID and the "update-source interface", on the core
> routers
> 	> (PE1)
> 	> is a loopback interface that belongs to the OSPF area 0.
> 	> The router ID and the "update-source interface", on the access
> routers
> 	> (PE6)
> 	> is a loopback interface that belongs to the OSPF area 10.
> 	>
> 	> After establishing OSPF adjacencies between all routers, the
> BGP
> 	> process
> 	> starts to establish connection, and this undesirable behavior
> happens:
> 	>
> 	> When the router PE1 (area 0) wants to establish a BGP session
> with
> 	> router
> 	> PE6 (area 10), the packet flow through all 100Mbps (purple
> arrow). When
> 	> the
> 	> router PE6 (area 10) responds, the packet flow through the
> 1Gbps
> 	> connection
> 	> between the core routers (red arrow). Every flow that needs to
> use the
> 	> LSPs
> 	> will do logically the same, causing out-of-order packets at the
> 	> network.
> 	>
> 	> I know that this is an expected behavior, as intra-area routes
> are
> 	> preferred
> 	> over inter-area routes, no matter what the link cost is.
> 	>
> 	> The question is: What solution do you guys think it's better
> for this
> 	> scenario, so that the packet flow goes always through the
> optimal path?
> 	> - Sham-links;
> 	> - Extended area 0 to one more hop;
> 	> - Change the "update-source interface" for area 10;
> 	> - Create small areas between the core and access routers
> 	> - Other solutions...
> 	>
> 	> We are planning to deploy a network with more than 200 PE
> routers in a
> 	> similar scenario, and I don't think that a single OSPF area is
> a good
> 	> choice
> 	> for us.
> 	>
> 	> Can anybody help with some advice?
> 	>
> 	> --
> 	> []'s
> 	>
> 	> Lívio Zanol Puppim
> 
> 	> _______________________________________________
> 	> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> 	> https://puck.nether.net/mailman/listinfo/cisco-nsp
> 	> archive at http://puck.nether.net/pipermail/cisco-nsp/
> 
> 
> 
> 
> 
> 
> --
> []'s
> 
> Lívio Zanol Puppim