i hav the following setup in design stage..i need to analyse the viability & best options for the
n/w.It has 2 PE's(metro-e) from different providers over MPLS cloud, corresponding CE's at our end hooked upto a L3 6500
series.Cloud is running BGP , i would like to load balance between the link so that it
achieves around 35-55% load on either paths from L3 towards either CE's ,using
either ospf/eigrp on my side towards the cloud and then provider would be
redistributing this with their bgp.
the igp(ospf possibly) would be run on the core switch.As usual CE's
for both links would be controlled by the providers running BGP.any ideas
on how to implement load sharing across the 2 links and also provide redundancy
in case of route failures from any PE to CE path.
Thanks for any suggestion/help in advance on this!
this should be possible if the two providers are providing you L3 VPN services they can carry OSPF routes inside their internal MP-BGP.
All OSPF routes of other sites will appear as O IA (OSPF inter -area).
If the costs advertised on other sites to th two providers are the same they should be presented as the same to your local CEs allowing the load balancing to happen in both directions.
All the complexity is on the service provider side.
If it was a single provider the results could be different depending on the preferences to the PE nodes.
With two different providers the load balancing depends on you choices: traffic sent to providerA will reach remote siteA on providerA link
Hope to help
Thanks Giuseppe for responding!
Actually both different providers are providing Metro-E links with 4M each and we'r planning that their CE's,(which we wont have access to)will be terminated on a L3 6500 s.w,and then the internal n.w follows.Since they wud have BGP from their PE's till CE's, we were thinking of using OSPF possibly on the L3 to kinda achieve loadbalance & also hav a redundancy/switchover in case the route from one path of CE to PE fails.
Any config eg. or any specific suggestions if this can be achieved by ospf.
Appreciate all your help!
Yes..both CE's would be hooked to PE's from 2 provider's.The situation here is the 2 CE routers would be coming from a single provider,so they should in all way be hooked together by ibgp & they wud be termined on 2 SVI on L3 & yes it needs to utilise both PE links in sort of loadsharing potential. & also any way to switchover in case any of the CE to PE routes fail.
Thanks in advance!
so the scenario is:
your C6500 here
the two CE routers can redistribute BGP into OSPF.
It is better to have them use OSPF O E1 (metric type O E1) with the same seed metric value.
In this way your C6500 router will build two adjacencies and it will install two default routes like:
O E1 0.0.0.0/0 [110, xx] via CE1
or whatover specific prefix the CE receive and send to your C6500.
It is your choice to decide if you want to put the two CE in the same vlan and so have them to build an OSPF adjacency between them or not.
If they come from two different providers I would keep them separated.
In this way you have load balancing and fault tolerance.
Hope to help
Thanks for the explaination..
But the scenario is like:
L3 6500 S.W
ya i plan to keep both on seperate vlans..any advantages/disadvantage in this?
so,am thinking of using max paths as ospf wud be running on L3 sw.and any ideas on how to facilitate switchover from one to another in case the route from one CE TO PE fails.
Thanks for all ur suggestions in advance!
OSPF has already a default of 4 maximum-paths so you don't need to change it in your scenario.
Actually when PEx - CEx link fails CEx stops receiving the best BGP path via PEx BGP next-hop.
So you can use a route-map made of two blocks
in first block you use
route-map generate-OSPF-routes permit 10
match ip next-hop 11
set metric type 1
set metric 50
access-list 11 permit host PEx-BGP-next-hop
if the link fails the BGP route is learned via another BGP next-hop (PEy or CEy doesn't matter ) CEx doesn't redistribute into OSPF the route
the C6500 will then have only one path via CEy.
Without the match ip address condition check CEx will always advertise a prefix regardless it is receiving it on PEx-CEx BGP session or via BGP session with CEy.
convergence time may vary depending on different factors (eBGP fast-external fallover and BGP per neighbor timers plus frequency time of redistribution checks)
Hope to help
thanks Giuseppe, this was gud info..got some more things to clear.
for ospf,i will hav to define say cost 100 on both the interface on L3 where the CE would come in ?
Got bit lost in the routemap-i understood that its to force the condition to be considered when the Pe-CEX link goes down..correct?
will help if the route map is more clarified alongwith the acl.
I already used this method to generate OSPF default routes if a BGP session is up.
the ACL is a simple standard ACL
access-list 11 permit host 10.10.10.1
here what is different is that we are not checking the prefixes but the next hop of the prefixes
so if in routing table there is
B [20/0] 192.168.5.0 via 10.10.10.1
this matches the route map and it is redistributed into OSPF.
if later session fails
another BGP prefix can be installed but with a different next hop and the idea is to track the next hop identity
Hope to help