Posted on July 15, 2015 7:53 am
 |  Asked by Victor
 |  5688 views
RESOLVED
0
0
Print Friendly, PDF & Email

Hi,

I wanted to know what is the advantage in using EBGP over IBGP in Spine and Leaf deployment. For example instead of having my Spines in one AS and then having each Leaf I deploy in it’s own AS and run EBGP why can’t I just run IBGP and use one single AS for both my Spine and all the Leaf I deploy. Issues such Next-hop unchanged will be corrected by making all my spines route-reflectors.

Thank you,

Victor

0
Posted by ma
Answered on July 15, 2015 1:57 pm

Hi Victor,

The first thing that comes to my mind is simplicity, and most likely you’ll see little faster convergence while using eBGP. But nothing stops you from running iBGP, if that is what you prefer. In that case I would advise the use of BFD for your BGP neighbors for better convergence.

Sharing some interesting articles which you may find useful:
Introducing data center fabric, the next-generation Facebook data center network
https://code.facebook.com/posts/360346274145943/introducing-data-center-fabric-the-next-generation-facebook-data-center-network/

Use of BGP for routing in large-scale data centers
http://datatracker.ietf.org/doc/draft-ietf-rtgwg-bgp-routing-large-dc/?include_text=1

Thanks,
-Michael

0
Posted by Piotr
Answered on July 15, 2015 10:22 pm

Hello Victor,

Like Michael wrote, you can use iBGP freely. One major concern that comes to my mind is scalability. As your network grows you will have way to many RR in your iBGP. Minor thing is as_path in your case will not give you any idea of actual routing path, compared to eBGP design.

Also, nice presentation from few years ago by Microsoft employees: https://www.nanog.org/meetings/nanog55/presentations/Monday/Lapukhov.pdf

Best regards,

0
Posted by Victor
Answered on July 16, 2015 12:45 pm

Hello both,

First of all thank you both for your responses. I just brought this question since I know a lot of deployments are using EBGP and never mention using IBGP. Regarding the fact that the more my network grows the number of RR would grow. But that would not happen because my only RR would be my Spine switches and I wouldn’t be increasing the number of Spines. Also each Leaf switch will be multihomed to each Spine switch. So the same routes will be reflected by each Spine switch. Also regarding loosing visibility to the AS_PATH that would not matter since in a IP Fabric deployment if all the switches are multihomed to each Spine switch from any point in the network it will only be three hops(three ASs).

Thank you,

Victor

 

 

0
Posted by ma
Answered on July 20, 2015 3:13 pm

Hello Victor,

If you prefer iBGP in your environment feel free you use it, I’m sure you will not be the only one running iBGP in a DC CLOS network. With regards to AS_PATH, some people use the AS_PATH to find the origin of route. And yes one could use communities to find out the same.

Just out of curiosity what would be your main arguments to pick iBGP over eBGP?

Best regards,
-Michael

0
Posted by ma
Answered on July 20, 2015 3:13 pm

Hello Victor,

If you prefer iBGP in your environment feel free you use it, I’m sure you will not be the only one running iBGP in a DC CLOS network. With regards to AS_PATH, some people use the AS_PATH to find the origin of route. And yes one could use communities to find out the same.

Just out of curiosity what would be your main arguments to pick iBGP over eBGP?

Best regards,
-Michael

Post your Answer

You must be logged in to post an answer.