MPLS Label Allocation in Aggregate Routers

arun kumar · ‎11-19-2007

Hi,

I ve a doubt in Route Aggregation with MPLS Label generation...

Suppose i have the following connectivity:

R1--R2--R3--R4--R5--CE2

--R6--CE1

R4 is having two connectivity one to R5 and R6 and aggregates the two Customer subnets into Single Subnet... For eg..

CE1 10.10.1.0/24

CE2 10.10.2.0/24

R4 aggregates two CE subnets into 10.10.0.0/22 and giving it to R1, R2 & R3.

As far as my understanding...

R4 will have the LFIB entry as Aggregate so that it looks into its Routing Table for sending the packets to either of the CE.

SO what would be the label value R4 generates and gives to R1,R2 and R3... Because in R4'r routing entry has two /24 but in R1,R2 and R3 routing entry has one /22.... (Will R3 and R4 exchange the labels since both the routers have different prefix lengths)..

How the labels are generated and exchanged between the routers and how the traffic is flowing with labels prepended..

Thanks in advance...

swaroop.potdar · ‎11-19-2007

Arun,

R4 cannot aggregate the CE1 and CE2 subnets as its a P router and has no knowledge of the CE subnets. As none of the routes from CE1 and CE2 originated through R4 from BGP nexthop perspective.

Although aggregation can be done for link addresses within the MPLS cloud (PE-P/PE-PE links) but not for the BGP nexthop loopbacks.

As this will result in LSP path break.

Again to apply this for your example, if R4 has received 10.10.1.0/24 and 10.10.2.0/24 prefixes as link addresses between R4-R5 and R4-R6 respectively then it can summarize it,

also since for traffic traversing the MPLS cloud the traffic will always be for BGP nexthops (the label swapping will be done based on the labels for loopbacks) so the label for the aggregate route wont come into consideration (untill and unless some traffic is destined to that range eg:10.10.0.0/22) which never can be the case as all customer routes are advertised via BGP (MPBGP or Ipv4 BGP) and never into IGP. So always the customer traffic is label switched even if you summarize the infra link addresses, but never summarize the bgp nexthop loopbacks.

HTH-Cheers,

Swaroop

raghavbhargava · ‎11-19-2007

so swaroop what i understand from ur last comment is perfectlly true..i.e-- the routes traversing thru the mpbgp (vpnv4) which are basically the customer routes are never aggregated nor can the P routers aggregate them coz they are only responsible for label swapping.the routes which are running b/w the pe-pe thru igp can be summarized or aggregated coz they ar only used for next hop reachibility.

Secondly incase the customer routes need to be aggregated only the pe routers can aggregate them.

Please advise??

raghav

swaroop.potdar · ‎11-20-2007

Yes Raghav, the PE routers only can aggregate CE routes apart from the CE's themselves.

Now a SP never summarizes CE routes as a best practise, at the most you can limit the number of routes per vrf as a part of your SLA.

HTH-Cheers,

Swaroop