Abstract

ABSTRACT Modern networking applications replicate data and services widely, leadingto a need for location-independent routing- the ability to route queries directly to objects using names independent of the objects ' physical locations.Two important properties of a routing infrastructure are routing locality and rapid adaptation to arriving and departing nodes. We show how these twoproperties can be efficiently achieved for certain network topologies. To do this, we present a new distributed algorithm that can solve the nearest-neighbor problem for these networks. We describe our solution in the context of Tapestry, an overlay network infrastructure that employs techniquesproposed by Plaxton, Rajaraman, and Richa [20]. 1 Introduction In today's chaotic network, data and services are mobile and repli-cated widely for availability, durability, and locality.1 This has lead to a renewed interest in techniques for routing queries to objectsusing names that are independent of their locations. The notion of routing is that queries are forwarded from node to node until theyreach their destinations. The location-independent routing problem has spawned a host of proposals, many of them in the context ofdata sharing infrastructures such as OceanStore [14], FarSite [3], CFS [12] and PAST [24]. To permit locality optimizations, it is im-portant that the routing process use as few network hops as possible and that these hops be as short as possible.Properties that we would like from a location-independent routing infrastructure include:

Citations