An analysis of naming conventions for distributed computer systems

Abstract

Name servers that collectively manage a global name space facilitate sharing of resources in a large internetwork by providing means of locating named objects. The efficiency with which the name space can be managed is strongly influenced by the adopted naming convention. Structured name spaces are shown to simplify name space management from both an administrative and system viewpoint. Formulae have been derived which allow one to quantitatively measure the effect of the distributed name server configuration on a given client's level of performance. In general, the cost of a name server query can be reduced by distributing replicated copies of name server database entries in a way that exploits the locality of clients' reference patterns.