2009/12/31 Sriram, Kotikalapudi [email protected] > > Follow up question from Charrie Sun: > Then this backup information can make the same storage and distribution > cost as to the original non-aggregated mappings, right? > > Response to the followup question: > KS: That is not the case. Only a small fraction of subnets are multi-homed > to multiple > ASs as you can see from slide 40 at link below (based on analysis performed > at NIST). > > http://www.antd.nist.gov/bgp_security/publications/ARIN_NetHandle_OriginAS_Analysis.pdf > Actually, in this slide many of the MOASs are due to _stale_ route > registrations at the IRRs! > So the fraction we are talking about is really small. > This is because, in reality, most of the time a subnet is originated by a > single AS > and it is the AS that is multi-homed to ASs of multiple upstream ISPs and > not the subnet. > Similarly, I think it would be rare that a subnet (or subsubnet) would > multi-home directly > to multiple ETRs. More often a serving ETR at lower-level would multi-home > to multiple > ETRs at higher-level in the hierarchy. The main savings in the proposal > is due to efficient handling of deaggregates that reside at another ETR > away from the ETR where bulk of the aggregated prefix resides. > > Here I'm confused about the separation granularity. Isn't the separation to separate stub AS from the core, but rather to separate subnets from their ASes? As there are often one AS for each subnet, what's the use we do so? The same question applies to ETRs.
Letong
_______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
