next-address question

rossonieri#1

Hi guys,

i have a few questions regarding MPLS TE,

using scenario :

CE01 ---> PE01 ---> CE02
---> PE02 --->


from CE01 - can my tunnel0 explicit-path name next-address use both PE01 & PE02 to reach CE02? to form a load balanced traffic?

or

should i use single tunnel0 or more tunnels to accomplish that?

or

it cant never be load-balanced?

any idea would be appreciated

thanks.

EdTheLad

You would need 2 tunnels.The next path attribute is used as a vector to determine which link the tunnel will be created on.The first tunnel would go CE01-->PE01-->CE02 it would need 2 next addresses specified the link between CE01 AND PE01 and the link between PE01 and CE02.
The second tunnel would go from CE01-->PE02-->CE02 which would also have two next-addresses specified i.e. link ip on PE02 and link ip on CE02.The metrics of these tunnels would need to be adjusted so that the path from CE01 to CE02 is equal over either tunnel.A "show ip route" on CE01 would show the equal cost routes with the exit interfaces being replaced with the tunnel ids.

rossonieri#1

hi ed,

sorry for this very late reply, and thanks for replying.
not good for the moment - still stuck.
cant post the output here since i'm keep forgotten to capture one.

all of the tunnel were up - but still the traffic is passing by tunnel0.
pending job for the moment.

and - do you have any experience implementing VPLS?
can you tell me exactly how it works?

i'm wondering whether it just like basic switch stacks around the block using multiple instances of STP?

TIA.

EdTheLad

Hey Ross, can you tell me is your load-balancing over vpls? or are you just testing the mpls to begin with.The load-balancing over the TE tunnels will be per destination so are you sending traffic to multiple destinations to dest? Do you see the two equal cost routes when you look at the routing table?
VPLS is an extension of EoMPLS, so a good start would be to read about EoMPLS first.On each PE you configure a VFI(virtual forwarding interface) instance, this VFI is then cross connected to remote PEs either manually or dynamically, the effect is you will have a mesh of VCs between PEs that is created using targeted ldp, so any traffic passing over this VC gets an additional mpls label.You can map one vlan to a VC or multiple vlans to one VC, so when a frame arrives on ingress the vlan id is examined and traffic gets mapped to the appropriate VFI as per your config.The frames get assigned an inner label for the VC and on egress towards the mpls core they get another label for the mpls forwarding, if you use TE you get another label.At the far end the labels get stripped and the frames are send out the ports which are associated with the receiving VFI.
An example, if you had 3 PEs, you could setup 3 VFI's, VFI 1 maps vlans 1-100, vfi 2 maps vlans 101-200 and vfi 3 maps vlans 201-300.When setting up these VFIs you will either manually specify the neigbor PEs or configure bgp autodiscovery either way you should have a mesh of VCs for each VFI.Unknown unicast traffic arriving on an ingress port with vlan 10 will be mapped to VFI 1 and sent across 2 VCs towards both neighboring PEs, i.e. it gets repliacted, know unicast traffic will be forwarded across only one VC as the mac-table will know which VC id to send on.As you can see broadcast traffic gets replicated so you must be careful on how much broadcast traffic you transmit.Vpls is used mostly to reduce the size of MST domains, bpdus get tunnelled through the vpls cloud.
Hope that helps, theres alot going on with vpls so its not so easy to explain.Its a bugger to troubleshoot as alot of labels to trace, also load-balancing is random as at the moment it uses a hashing algorithm of the inside label, since the inside label is system generated its not possible to manipulate.

rossonieri#1

hi Ed,

no i'm not trying to do a load-balancing over vpls - that was only my customer project for their mpls (the load-balancing). i'm just thinking to extend my service to handle vpls implemetation.

and like i said before - i have CE01 with 1 tunnel0 (say 1.1.1.1/32) go out to s1/0 and s1/1 to reach remote CE02 tunnel0 (2.2.2.2/32), in the routing table it only has 1 OSPF route to remote CE02 tunnel0 - could this be a mistake?

that is why i ask whether i need to create 2 tunnel interface on both CE01 and CE02?
like :
CE01 tunn0 1.1.1.1/32, tunn1 1.1.1.2/32, and
CE02 tunn0 2.2.2.2/32, tunn1 2.2.2.3/32

so i created 2 static route using these tunnels to form a load-balanced traffic? (just for temporary replacing the OSPF - to see whats happening)

then i did create 2 tunnels on each side - up and up - but the result is that i keep getting only 1 tunnel forwarding the traffic to remote end.

but - i'm desperately need a break - so i'm pending this job

ya - i've begin reading VPLS - and gets me a little bit blur now - since its implementation not too far from mpls network

by the way - thank you very for the big explanation Ed. Great helps, thanks.

and...

Wishing you a very Happy New Year for all of you guys TE members