ABSTRACT: In this paper we present a new algorithm for retrieving and updating data from a distributed relational data base. Within such. a data base, any number of relations can be distributed over ' any number of sites. Moreover, a user supplied distribution criteria can optionally be used to specify what site a tuple belongs to. The algorithm is an efficient way to process any query by f'breaking'l the qual-ification into separate "pieces " using a few simple heuristics. The cost cri-teria considered are minimum response time and minimum communications traffic. In addition, the algorithm can optimize separately for two models of a communi-cation network representing respectively ARPANET and ETHERNET like networks. This algorithm is being implemented as part of the INGRES data base system.