Hi, Lixia Thank you for your comments. Please check my justification inline.
> Date: Mon, 18 Jan 2010 23:27:45 -0800 > From: Lixia Zhang <[email protected]> > Subject: [rrg] a quick critique on 2-phased mapping > To: [email protected] > Message-ID: <[email protected]> > Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes > > This is a simple idea on how to scale mapping. However personally I > feel the design is too incomplete to be considered a serious input to > RRG. Take the following 2 issues as example: > The design is simple indeed but not incomplete. It focuses on mapping mechanism and can work with many Id/Loc separation proposals(e.g. LISP) nicely. If you need map and encap/rewrite as a whole package I can supplement the rest part, but I think the supplementary can hardly attribute new ideas other than LISP/IVIP etc. I am afraid of distracting readers' attention from too many complicated details. I believe the scalability of mapping in the ID/Loc separation scheme is extremely important and relatively independent of tunneling/rewriting techniques. The routing scalability problem will be shifted to mapping system if the scalable routing presumably acquires mapping beforehand. > First, in this 2-phase scheme, an AS is essentially the unit of > destinations (i.e. sending ITRs find out destination AS D, then send > data to one of of D's ETR). This does not offer much choice for > traffic engineering. > This mapping scheme will not influence intra-domain TE within an AS. If an AS expects some inbound traffics to certain prefixes come in from desired ETRs, it is in the case of interdomain TE. 2-phased mapping guarantees optimized route in the core(BGP routing with RLoc). If the destination AS D needs to enforce its interdomain TE objective in the core, there will be a contradictory between the core and edge. Therefore any gain of the "edge" means undesired cost of the "core". Any way if the AS concerns more about the interdomain TE, the ETRs can be configured to redirect special traffic to a specific inlet of the AS. But this configuration is not an essential part of the mapping system. Generally speaking it is reasonable to keep the granularity of mapping on AS level, which will reduce the mapping dynamics efficiently. > Second, there is no consideration whatsoever on failure detection and > handling. > > Lixia Since the 2-phased mapping scheme takes the advantages of DNS and BGP at the phase I and II respectively. Therefore there is no special failure detection/handling design besides DNS and BGP inherent failure recovery mechanism. Even though the mapping information in the DNS server or mapping cache in the ITR can be obsolete and causes routing failure, ETRs will send destination unreachable message to ITRs to triggre immediate mapping queries. Or when an ETR is down, the BGP update will inform the ETR failure to ITRs. ITRs will choose alternative ETRs to the destination AS or send icmp packets to sources. The ETR reachability detection and routing failure recovery can be implemented variably from case to case. However it will not have substantial impact on the functioning of 2-phased mapping scheme. _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
