I am a consultant assisting many customers connect to MPLS providers like (AT&T, EQUANT, etc). My question is that why do providers place a restriction on using BGP only to exchange routes with the PE router. Cisco IOS supports RIP, EIGRP, OSPF. What is so special about BGP that many providers have setup a policy to only BGP as far a peering goes with the CE.
BGP is a natural choice for SPs when it comes to connecting customers to their infrastructures. IGPs are not.
BGP easily allows them to enforce policies at the edge of their network.
Hope this helps,
i appreciate your response. But to have a policy in place is difficult to understand. Many organizations are running eigrp or ospf or rip and now they have to deal with redistribution and other complimated conversions just to interface with the provider.
I understand when you say that they prefer, but, to have a policy in place that it is BGP or static makes me wonder, atleast that is what AT&T told me when i told them i wanted to use EIGRP to interface with their PE router.
I think you need to look at the difference between Exterior and Interior routing protocols. BGP is THE best choice to connect a customer to the internet. All of the protocols you mention are IGP's and aren't really designed to handle the needs of routing out on the internet. Here is a quick little article I found for you.
As I said, it makes it easier for the SP not necesseraly for the customer ;o)
If the customer is adamant about preserving their IGP when connecting to the SP and that same customer has a substantial number of sites, in some cases it might influence the SP, which might change its policy regarding the protocols customers can use.
I have seen some providers supporting OSPF and EIGRP along with BGP and static just because they know how many customers are running these IGPs in their network and that it might be appealing for customers to sign with them rather than the competition. It becomes a competitive diffentiator.
Bear in mind that for the SP, there is an operational cost linked to supporting these additional protocols between the PE and the CE.
Hope this helps,
Some of the points i could think of for that especially for MPLS-VPNs
1) EIGRP maybe a restriction as some providers may have Juinper on the edge
2) In case of inter-AS MPLS, type of interconnect plays a major role. An option A NNI creates lot of problems as the LSP is broken at the NNI
3) Easier to do traffic enginnering with BGP using communities etc certainly tougher in IGPs.
4) Finally I would say the question of IGP, not many providers are comforatble extending customer IGP to their network.
1) this is a good point but it is really the SPs problem. If offering EIGRP as a CE-PE routing protocol brings you lots of entreprise customers because EIGRP makes sense to them then the SP might reconsider the devices they use as PE routers.
I have worked with one major SP in the US offering EIGRP for this exact reason. It is simply a commercial issue.
It was explained to me that the SP exponentially increases the number of routes their routers must maintain, as all routing protocols on their PEs have to be redistributed via BGP. So, if an SP takes in 50 routes from a customer, that PE actually maintains 100 (50 in the native IGP, and 50 via BGP).
BGP is often the preferred CE-PE routing protocol as it is the most controlable (=secure) option for the provider.
CE-PE BGP may also have scaling benefits compared with IGPs
one of the issues a SP has is to protect the PE, which is used by many customers. This means basically two things: security and resource protection (ok, could be seen as part of security as well).
Simplified: security is independant of routing protocols (access-lists).
Resource protection is dependant on the routing protocol. Assume f.e. a customer connected to a PE with OSPF. What will the memory and CPU load be the customer poses to the PE? Assume further the customer redistributes the complete Internet routing table as external routes into OSPF.
Then 170000 external routes will be in the OSPF database consuming CPU and memory on a PE. This could eventually crash the PE - not acceptable.
You can restrict the number of prefixes inserted into the vrf routing table, but afaik not the ospf database.
With BGP you simply define a maximum prefix limit and terminate the session when the customer exceeds it.
(Hm. ... a way to make a little extra money for the SP? Define a max of 5 routes per VRF and raise it for lets say 10$/month per route? ;-)