Client-site query extensions

Abstract

We explore the execution of queries with client-site user-defined functions (UDFs). Many UDFs can only be executed at the client site, for reasons of scalability, security, confidentiality, or availability of resources. How should a query with client-site UDFs be executed? We demonstrate that the standard execution technique for server-site UDFs performs poorly. Instead, we adapt well-known distributed database algorithms and apply them to client-site UDFs. The resulting query execution techniques are implemented in the Cornell Predator database system, and we present performance results to demonstrate their effectiveness. We also consider the question of query optimization in the context of client-site UDFs. The known techniques for expensive UDFs are inadequate because they do not take the location of the UDF into account. We present an extension of traditional 'System-R' optimizers that suitably optimize queries with client-site operations.