I really have to disagree with this view that IGPs will wither away. Yes, in the extreme case, when one is using a logically singular controller (possibly a collection of controller sharing state using a proprietary mechanism, but appearing as one) to control the entire autonomous system (enterprise, data center, ...) then yes, you can dispense with the IGP.
But, if you expect the solution to scale (to support large data centers, large enterprises, or autonomous systems the size of those preferred by many Internet Service providers), then you are going to need multiple controllers, and those controllers will need to talke to each other. That is exactly what the existing IGPs are designed for. So I would not forsee the demise of the IGP. (And, as a side-effect, this gives an easy way to handle the necessary inter-working with existing routing devices.) Yours, Joel > -----Original Message----- > From: openflow-discuss-boun...@lists.stanford.edu > [mailto:openflow-discuss-boun...@lists.stanford.edu] On > Behalf Of Wes Felter > Sent: Monday, May 20, 2013 12:18 PM > To: openflow-discuss@lists.stanford.edu > Subject: Re: [openflow-discuss] SDN & Routing Protocols > > On 5/18/13 9:24 PM, Farhad Ibrahim wrote: > > > What i specifically want to know is: Will future SDN networks need > > OSPF/EIGRP? Does SDN in its pure form mean the end of OSPF/EIGRP? > > Yeah, with SDN you either don't need any routing protocol at > all or you can invent a new more "modern" routing protocol > based on Paxos. > > -- > Wes Felter > IBM Research - Austin > > _______________________________________________ > openflow-discuss mailing list > openflow-discuss@lists.stanford.edu > https://mailman.stanford.edu/mailman/listinfo/openflow-discuss > _______________________________________________ openflow-discuss mailing list openflow-discuss@lists.stanford.edu https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
