Aggregation within ISP network

Terence Payet · ‎01-18-2015

Hi expert,

I need your advise on a new deployment.

I currently work in a small ISP, and we currently have over 20,000 customers.

What we want is to seperate our core from border gateway, i.e. we want to use BGP on the border ASR and ospf on the core and BRAS, so basically we want to advertise some sort of aggregate route to our border router (see attached diagram).

So my question is. In an ideal ISP deployment:

Can we send default routes from BGP to IGP.
We will need to configure iBGP between Border and core egress interface right?
Do you have some sort of deployment config or examples on how to do this.

Thanks to advise.

Steve

Vasilii Mikhailovskii · ‎01-19-2015

Hello.

1. If you configure your core with IGP only, keeping BGP on ASR only, then you won't have any option for outbound load-balancing over uplinks.

2. You need iBGP only if you need any consistent decision across the ASRs about BGP best path (or if your AS is a transit)... if not, you may just advertise default gateway into IGP (default-information originate metric ...) and tune IGP metric to prefer one over another.

3. You may find BGP basics here - http://www.cisco.com/c/en/us/td/docs/ios/12_2/ip/configuration/guide/fipr_c/1cfbgp.html

---

Q:

- are you going to run eBGP with your customers?
- if yes, are you going to accept provider-independent prefixes?
- are you going to provide your customers any provider-dependent prefixes? What will be the routing protocol (again - eBGP)?
- if yes, how many prefixes are you going to "sell" to the customers?
- are you going to run MPLS in your core?

Terence Payet · ‎01-19-2015

Thanks Vasilii.

Please find below answers for your questions below:

- are you going to run eBGP with your customers? = YES

- if yes, are you going to accept provider-independent prefixes? = YES

- are you going to provide your customers any provider-dependent prefixes? What will be the routing protocol (again - eBGP)? In the future YES, and eBGP will be use.

- if yes, how many prefixes are you going to "sell" to the customers? = Don't know yet

- are you going to run MPLS in your core? No. MPLS is on a separate segment.

What we want is to redesign the whole ISP network. Currently we have OSPF running on the ASR and core from which OSPF is redistributing customer prefixes into BGP. We are only accepting default routes from the providers, but we want to receive the whole routing table so proper traffic engineering can be implemented.

So your thoughts on best practices is highly appreciated.

Thanks,

Steve

Vasilii Mikhailovskii · ‎01-20-2015

Hello.

If you receive full view, what device (do your plan) is to make routing decision about the best path?

I thought you were not going to enable BGP in the core... so MPLS is probably required here.

Are you going to inject customers' (PD and PI) prefixes into IGP?