There's been a lot of research on the relationship between overlay topology and physical network topology. I would start with Pastry and go from there.
At a high level, there's a tradeoff between optimizing performance by ensuring that physically close peers are close in the overlay and optimizing fault tolerance by minimizing information loss when a network partition occurs (by ensuring that data replicas are not close in the physical topology). Bruce xianghan.zheng wrote: > Hello, > Is there some proposal that suggest the physical closed peers will have > the similar identifier? I think this is necessary. Somebody have some > suggestion? > > Best Regards, > Xianghan Zheng > > > ------------------------------------------------------------------------ > > _______________________________________________ > P2PSIP mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/p2psip _______________________________________________ P2PSIP mailing list [email protected] https://www.ietf.org/mailman/listinfo/p2psip
