Resource location in very large networks

Networks such as the Internet and the telephone system are examples of ubiquitous large networks. While such networks are quite common, what is not so common is completely scalable, non-hierarchical naming that is independent of the entities location or affiliations. The problem of providing a unique, logical name for each nameable entity that can be kept immutant over the lifetime of the entity has been studied in many contexts. However, most results are not scalable for really large worldwide networks. In this paper we first motivate the need for such naming schemes and then explore the possible name translation solutions. We develop a set of pragmatic criteria that a solution has to provide. The scenario we use is a hypothetical, but desirable naming scheme for telephony. In our scheme, each resource (human, telephone or computer) has a unique name that is neither mutable, nor dependent on any hierarchy of domains or geographical boundaries. We then provide an engineering solution for resolving the location of a resource given its name. This solution has to be efficient, scalable and fault-tolerant. The solution uses several disjoint techniques such as caching, physically hierarchical servers and hashing. We show how our scheme meets our goals and argue its feasibility.