cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1698
Views
0
Helpful
5
Replies

MPLS load balancing

kumarmh91282
Level 1
Level 1

Hi All,

 

Thanks in advance,

 

We have a Dual MPLS connection from our Data Center from Remote location as shown in the attached picture.

 

currently we are using one mpls as Active and other as standby (Using Path Prepend)

 

CE to PE is EBGP connection 

and CE to Core Router is IBGP connection as shown in picture

 

Please guide/suggest us the best practices to load balance the MPLS links and efficiently use the bandwidth.

 

Thanks

5 Replies 5

Joseph W. Doherty
Hall of Fame
Hall of Fame

It might possibly be as simple as config "maximum-path 2" as the number of ECMP routes allowed in your BGP (for your iBGP) on both your internal (i.e. those routers with connections to your CEs) routers.

Giuseppe Larosa
Hall of Fame
Hall of Fame

Hello @kumarmh91282 ,

if you want to achieve load balancing you would need to remove AS path prepending towards current secondary provider and in your core router you would need under router bgp

eibgp maximum-paths 2

 

the normal command applies only to eBGP sessions

note that depending on platform and IOS XE version the suggested command may be:

not availavable

supported in address family ip vrf <vrf-name>

supported under address family ipv4 unicast

 

Hope to help

Giuseppe

 

 

@Giuseppe Larosa makes an excellent point about the "particular" form of "maximum-paths" needed to work with iBGP.  If you're in the situation where it's not supported, you might also consider using perhaps eBGP between your core routers and CEs or perhaps an IGP (maximum-paths 2 more likely to work, with the former [correct Giuseppe?] and the latter, generally with Cisco, defaults to maximum-paths 4).

Giuseppe raises another point, if doing something like prepending, out, you'll need to adjust that.  Ditto if using any other "traffic engineering" to "steer" ingress traffic.

Also, depending on how your two MPLS vendors implement their internal (private) BGP topology, you may see a difference there too you need to adjust for.  (Years ago, I worked at a company doing MPLS/BGP/VPN, internationally across two different vendors.  The one presented the whole world as one AS, the other had regional ASs.  So the former, for inter-region communication, always appeared to have less "hops" on the one vendor than the other.)

Lastly, if your really want to use the two paths "efficiently", then you want to run something like Cisco's PfR on your CE routers.  Maximum-paths just, more-or-less, round-robins flows to "equal" cost destinations.  PfR dynamically moves flows about to load balance actual utilization.  It can also do this unequally.  It can also deal with "impairments" on one provider and shift to another.  It can also shift flows based on QoS considerations.  For example, you only show two sites, but if you had three, or more, how do you insure two sites sending to another, not only load balance their links to the MPLS clouds, but the egress on the links at the destination site?  PfR can do that too.

Hello @Joseph W. Doherty ,

yes in case of lack of support of the correct command to make the core device able to support iBGP multipath a possible option is to use an IGP between core device and the two border routers implementing appropriate redistribution of BGP into the IGP of choice.

 

In the case of IGP the maximum paths commands default is 4.

 

Again, this time at IGP level the two Border routers have to inject in the IGP external routes with:

same type  ( applies to OSPF O E1 are preferred over O E2 and recommended when there are multiple ASBR nodes advertsing the same set of prefixes)

same seed metric

the IGP metric to the two Border routers must be the same.

 

Please note that in each Branch site instead of redistribution of BGP routes advertising a default route in the IGP with same parameters as listed above can be enough.

The difference for OSPF you need the command

router ospf 10

default-information originate

 

on each branch WAN/border routers.

 

Hope to help

Giuseppe

 

Hello
Can you confirm if CE to CE connections have the same link BW or do they differ as if they do differ in cost then you would probably need to use unequal cost load sharing-  bgp-dmzlink-bw  and for that costing to reach the IBGP peers (internal dc ibgp rtrs) then you would need to append neigbour dmzlink-bw and maximum-path-ibgp


Please rate and mark as an accepted solution if you have found any of the information provided useful.
This then could assist others on these forums to find a valuable answer and broadens the community’s global network.

Kind Regards
Paul
Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: