BGP load sharing MID attribute

[R Manjunatha]

I have two hub routers, R1 and R2. R1 is the primary router for the office, and R2 is the backup. R1 is connected to 8000 offices through Flex VPN. My requirement is to share the load between the two hub routers, R1 and R2. Currently, I am using the BGP attribute MED to route all the traffic primarily to R1

R1

Rrouter 65020

bgp router id 10.10.10.10

bgp listing range 192.168.20.0/24 peer-group office

bgp listen limit 10000

address-family ipv4 vrf office

neighbor office peer group

neighbor office remote-as 65010

Neighbour office route-map office_BGP_OUT out

route-map office_BGP_OUT permit 10

set metric 100

R2

Rrouter 65020

bgp router id 20.20.20.20

bgp listing range 192.168.20.0/24 peer-group office

bgp listen limit 10000

address-family ipv4 vrf office

neighbor office peer group

neighbor office remote-as 65010

Neighbour office route-map office_BGP_OUT out

route-map office_BGP_OUT permit 10

set metric 110

[paul driver]

Hello

---

@R Manjunatha wrote:
Could you please provide the sample configuration of how to send traffic to 4000 office to R1 and 4000 office to R2

---

MED is a lower preferance bgp attributue when comparing the bgp route path selection proecess, more preferred attributes or features can be used to traffic engineering like condictional route advertsiment, however can you elaborate your OP please as Flexvpn just like dmvpn creates dynamic tunnels between communicating spokes, so do you mean routes that the hubs receive externally that is then advertised to all spoke rtrs.

[R Manjunatha]

Basically, this is a hub-and-spoke communication model. My requirement is to have 8000 offices and two HUBs, Hub1 and Hub2. I need to distribute the offices between the two HUBs, with 4000 offices connecting to Hub1 and the other 4000 offices connecting to Hub2. Which BGP attribute should I use to route the traffic between the two HUBs, and can you provide a sample configuration?

[Giuseppe Larosa]

Hello @R Manjunatha ,

you would like 4000 routers to build and mantain dynamic IPsec tunnels with Hub01 and the other 4000 sites to use Hub02 to build and send user traffic.

A possible approach is to use BGP standard communities.

Hub 01 can add a BGP community of 65010:1000 to all advertised prefixes and Hub02 a BGP standard comunity of 65010:2000.

spoke routers will then use the BGP community value to rise the local preference of routes received by Hub01 or Hub02.

you will need to send-community in all eBGP sessions between Hub and spokes

Otherwise you can directly rise the local preference per eBGP peer on each spoke site using

negihbor x.x.x.x local-preference 200.

Half offices will rise the local-pref to Hub01 IP address the others the local-pref to Hub02.

This means changing the configuration of all spoke devices using two different configuration templates.

Edit:

the BGP community value can be used on the Hub side to decide what prefixes received from spokes should have the local preference increased in a coordinate manner with settings on the spoke routers.

The spokes that advertise their local prefixes with attached a BGP community of 65010:1000 will be served by Hub01 ( increasing local pref to 200 inbound ) .

The spokes advertising their local prefixes with attached a BGP community of 65010:2000 will be served by Hub02 ( increasing local pref to 200 inbound)

when the two hubs routers exchange routes the higher pref prefixes are installed as best path in iBGP in each of them.

Hope to help

Giuseppe