cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
541
Views
15
Helpful
8
Replies

IGP protocol en Carrier Ethernet Network

Hello and happy new year everyone. I'm working on #CarrierEthernet network with +75 #Cisco ASR/NCS running #EoMPLS #LDP, #OSPF, #L2VPN I wonder if I have to migrate OSPF to #BGP or other IGP as our network is growing fast, currently all Label Switch Router (LSR) are running OSPF into same area. Any suggestion or experience will be appreciate. Thanks a lot.
8 REPLIES 8
Richard Burts
Hall of Fame Guru

We do not know much about your environment. And perhaps if we knew more it might change our advice. But based on the little that we know at this point I would say that OSPF was a good choice for routing protocol. I do not see anything in what you describe that would motivate a change to BGP. If you are communicating with 75 layer 3 devices exchanging routing advertisements in a single area some people might suggest thinking about dividing the network so that there were several areas. I do not know enough about your topology and environment to have an opinion about whether multiple areas would be an improvement or not. Is there any basis for grouping some devices together and being logically separate from some other devices? If so multi area OSPF might be a good suggestion.

HTH

Rick
Giuseppe Larosa
Hall of Fame Master

Hello @Landry Hermann SAMBA ,

before changing IGP or moving to BGP (that may require an underlying IGP by the way if i BGP)) you need to consider the following about OSPF scalabilty:

- one thing is how many routers can be supported in the same VLAN/broadcast domain I have seen up to 50

- another thing is now many routers can be suppported in a single OSPF area and this number ca be much greater like few hundreds. Tests of this type have been reported here in the forum by Russ White.

 

So a possible suggestion to grow keep using OSPF single area is to use CARRIER ethernet underlying with mutiple VLANs this should solve your issue using subinterfaces / service instances / BDI.

Of course scalability means that each emulated VLAN will have a different pair of DR and BDR using ip ospf priority command.

Warning: uptime dictate OSPF DR and BDR (there is no pre-emption of DR) after the wait time expires , so you may need to be careful to be able to make effective the use of

ip ospf priority

Cisco defaults to 1, 0 means not eligilble the higher value is better.

You may need  maintenance window(s)  to shut current DR /BDR to have load sharing to work.

H-VPLS can be another way to achieve greater scalability Hierarchical VPLS:

 

Hope to help

Giuseppe

Hello and thank you for your reply

I want to give more information about how our network for carrier ethernet is setup.

All ASR and NCS are connected in point-to-point (/31) for core interfaces, OSPF is running on core interfaces, there is no DR/BDR in this case. All interfaces core are in the same OSPF Area. we use LDP for label distribution. This network deliver services to customers like E-ACCESS, E-LINE, E-TRANSIT.

As we have more that 75 nodes (ASR and NCS) into one area OSPF (0.0.0.0), I have read cisco press book that suggest or recommend to not have more than 50 routers into a area https://www.ciscopress.com/articles/article.asp?p=1763921&seqNum=6 I wonder if I have to redesign our network with backbone area and others area, I can see also if I redesign we will have some constraints to our manner to connect new PoP (ASR/NCS) to our network, currently with all equipment into the same OSPF area (Backbone) we can connect new PoP as we want just put the new interconnexion link into area 0.0.0.0 and enable MPLS and LDP.

Please let me know if there are some suggestion.

Thank you.

Hello @Landry Hermann SAMBA ,

>> All ASR and NCS are connected in point-to-point (/31) for core interfaces, OSPF is running on core interfaces, there is no DR/BDR in this case. All interfaces core are in the same OSPF Area. we use LDP for label distribution. This network deliver services to customers like E-ACCESS, E-LINE, E-TRANSIT.

 

OK I don't see any issue here the limit reported in the link from Cisco Press  is taken from the first edition of Halabi "Internet Architectures" dated second half of 90s so the limit of 50 routers in a single area is not true anymore.

 

You say you are using LDP , do you use also RSVP TE in that case the mpls TE traffic.egineering area is only one and it is area 0 in standard implementation.

Do you use also MP BGP for auto discovery of VPLS ?

 

Edit:

https://wiki.mef.net/display/CESG/E-Transit

 

I see that you provide point to point services E-transit is a connection between two E NNI.

LDP is is enough , RSVP TE could be used for providing MPLS Fast Reroute protection of each pseudowire but your customers may be using ethernet OAM to detect faiulers in a timely manner.

 

Hope to help

Giuseppe

 

Hello @Giuseppe Larosa 

Thank you for your reply.

Only LDP, OSPF are running in our carrier ethernet network, MP BGP and RSVP TE are not used for the moment.

As our network will grow fast with more ASR/NCS Cisco and all those nodes are into the same OSPF area (0.0.0.0), I start think if there was a limit of in term of node into one area OSPF, so do you have some reference or guideline which shows that there is no limit in having many routers (+200 for example) into the same area.

 

Best regards

 

Hello @Landry Hermann SAMBA ,

a good reference is the following

https://www.ciscopress.com/store/definitive-mpls-network-designs-9781587051869

 

One case study is actually  the TIM MPLS backbone ( I had been working on in years 1999-2006)  backbone unfortunately the book is not sold anymore.

 

So you can find it here in attachment

 

Hello @Giuseppe Larosa

Thank you for your reply, I'll read the reference MPLS guide you share with me.

If any concerns, I'll share in order to continue the discussion.

Thank you.

 

GEORGIOSBEKOS
Beginner

Hello  Landry Hermann ISIS would be an option too to consider with BGP.

 

Do you need also MP BGP for auto discovery of VPLS ?

 

https://www.cisco.com/c/en/us/td/docs/iosxr/ncs5000/vpn/63x/b-l2vpn-cg-ncs5000-63x/b-l2vpn-cg-ncs5000-63x_chapter_01001.html

 

Can you explain why you need to migrate , there are many valid designs but all depends of what you need to accomplish or future problems that you want to overcome.

 

Best regards,