>The size of the network does have a role when choose between RR and Confeds. >The BGP desgin for a POP is quite different than for the backbone, this is >true no matter the size of the ISP.
Well, yes. If the POPs have different policies, confeds may be useful. Otherwise, I think you'll find most ISPs use a hierarchy of RR's, with an intra-POP RR acting as a client of a higher-level RR closer to the core. Depending on your aggregation strategy, the POP may be a single IGP routing domain, or a non-backbone area of ISIS or OSPF. Large POPs may even contain multiple areas. The core itself may or may not contain BGP; it may use an IGP and MPLS. >POPs are better place for RR where the backbone is better off with something >else. >If it is a purely dialup POP, you do not even need BGP within the POP, only >the routers connecting to the regional or main backbone need to advertise >your routes. Depends what the dialup POP is doing; there are lots of possibilities including having to be a L2TP proxy for regulatory reasons. >A typical design could be having RR and IGP non-backbone area in each POP, >making every POP in a Confeds and runing IGP backbone area across the whole >BGP backbone network. I'm really not picturing the sort of policy that would call for this. > >Just my .02 > >Kent Yu >""Gregg Malcolm"" wrote in message >[EMAIL PROTECTED]">news:[EMAIL PROTECTED]... >> Good question. I don't claim to be an expert on BGP and will not comment >on >> the advantages and disadvantages of RR/Confeds in large/small >environments. >> I've only config'ed BGP once in a prod net. Seems to me tho that there is >> an important distinction between the two as they might pertain to the lab >> test. Both allow the use of weight and local preference (IBGP). Confeds >> allow the use of MED since the connections between the confeds is EBGP. >You >> could also config multiple RR's (clusters) to allow the use of the MED >> between the clusters but to me, confeds are easier. I'm sure that there >> are other reasons to use one or the other as they relate to the lab, but I >> haven't run into it yet. >> >> Maybe I'm wrong, but this is my take on RR's/confeds. Would love to hear >> comments to enlighten me. >> >> ""Kane, Christopher A."" wrote in message >> [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... >> > I'm currently tearing apart BGP as part of my IE studies. It's not too >bad >> > since I come from a Network Service Provider background. But, I have run >> > into a conflict in regards to RRs vs. Confeds. I probably don't need to >> > straighten this out for the Written but when it comes to the lab I'd >like >> to >> > know which route to go down. I have no idea how the lab poses it's >> topology >> > but if given the requirement to configure a simulated "large" network >and >> > then having to choose whether to implement RRs or Confeds I wonder which >> one >> > Cisco prefers. I'm assuming that as part of the lab, the idea is to >create >> > solutions that work and in doing so, solutions that are as simple as >> > possible and as short as possible. >> > >> > I'd like to hear comments about the pros and cons of each option in >> regards >> > to how Cisco might prefer to see implementation. Meanwhile, I'm going to >> > review all available case studies on CCO. >> > >> > Thanks, >> > Chris Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=29996&t=29968 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
