Client/server based statistical computing

TL;DR: In this article, the authors propose a client server architecture for statistical computing based on TCP/IP, which allows for the development of specialized client programs for particular tasks and allows the possibility to connect various client programs via a TCP-IP connection to a powerful statistical engine.

Abstract: We propose a client server architecture for statistical computing. The main feature of our approach is the possibility to connect various client programs via a TCP/IP connection to a powerful statistical engine. This offers the opportunity to include the statistical engine into a number of software packages and to empower the user of these packages to access a modern statistical programming environment. It also allows for the development of specialized client programs for particular tasks. TCP/IP permits a client/server connection with the client and server running on different hosts (remote host) as well as running both applications on the same computer (local host). To have a large flexibility we suggest adding a middleware program managing the communication between Server and Client. This avoids the need to implement TCP/IP communication methods on the server side. The paper provides an overview of the desired environment and illustrates the general structure by the implementation of the XploRe Quantlet Client and XploRe Quantlet Server.