Vpnv4 Load-Balancing with Dual Home CE/Access-Router in IPRAN/Seamless MPLS Intra-AS sceneario

[Wizzle1]

Hi,

I am facing a problem of load-balancing problem in IPRAN / Seamless Intra-AS design.

As i tried to present in diagram,

1- Site Router is dual homed to 2 RAN Aggregator which act as L1-RR (RR for all Site Routers)..These L1 RR / AGG routers are Next Hop Self.

2- L1 RR are then client of L2-RR (on top).

Problem is that for all VPNV4 prefixes always AGG-01 is best route selected on PEs due to Lowest Router ID (other BGP attributes are same) and so all downlink traffic from PEs towards AGG follow just AGG-01. I have tested below solutions & limitations:

1- Route Aggregation on AGG, it has some administrative challenge (from Radio) so can't do it .

2- BGP Multipath command to install 2 Best Next Hops for VPNv4. We have some route-leaking, due to which, it will create problem for other traffic in same VPN.

What else solution i can try to have Next Hop of both AGG installed on PE so i can have load-balancing. If i missed any information, kindly inform and I'll share.

If my question is very basic or inappropriate, I apologize. 

[pigallo]

Hi,
scenario is not that clear, anyway you could use SR-TE with anycast prefix-SID to load balance traffic towards both PEs.

Kind regards.

[LinhNguyen15326]

i think just config next-hop self for bgp family ipv4 and use bgp-lu for singnaling .

In scenario IP RAN use MPLS TE for optimize traffic . 

[CCIE28]

Not very clear, but have you thought about changing the values of RD, making the routes for the same prefix different and thus tricking the PEs to install both routes?

[asiergd]

Hi Wizzle1,

about your question:

"

Problem is that for all VPNV4 prefixes always AGG-01 is best route selected on PEs due to Lowest Router ID (other BGP attributes are same) and so all downlink traffic from PEs towards AGG follow just AGG-01.

"

In other words you stated that:

BGP best path selection process at PE1 and PE2 for vpnv4 prefixes advertised by AGG-01 and AGG-02 uses the lowest router id (AGG-01 better than AGG-02) as tie braker.

If this is true it is probably that you are advertising AGG-01 and AGG-02 loopback prefixes as external type 2 in ospf area 0 (External type 2 is the default type when you redistribute prefixes (connected, static, other protocols, etc...) in ospf).

Of course the loopback prefixes what I am talking about are the loopbacks used by AGG-01 and AGG-02 at vpnv4. These loopbacks will be the next-hop of vpnv4 prefixes at PE1 and PE2.

External type 2 routes in ospf does not "add" the ospf cost at each hop, they remain with the same metric value. Thus the metric to AGG-01 loopback and the metric to AGG-02 loopback will be the same in PE1 and PE2.

Please check the routes of AGG-01 loopback and AGG-02 lopback in PE1 and PE2 in order to verify this point.

If the statements are correct there are multiple solutions that you can do to compute the best route on PE1 and PE2 based on the closest AGG node, here go one of them:

 ->Redistribute AGG-0X loopback as external type 1 in ospf. This type of ospf route will add the ospf cost at each hop, thus PEs will compute the best path based on closest AGG node.

Of course you can advertise loopback interfaces as intra-area ospf prefixes, but if you are using external prefixes perhaps is good idea maintain these prefixes as external prefixes.

Regards and good luck!