Client-server applications on distributed-parallel and distributed-object platforms

Client-server applications have been traditionally implemented on distributed-object platforms. Although very little research has been done on developing client-server systems on distributed-parallel platforms, these platforms offer advantages which stem from parallel computing. This thesis compares two specific platforms, PVM and CORBA, for implementing client-server applications and presents a client-server system which is used to compare both platforms. This system consists of networked, client and server tasks which search the entire chess game tree for a 'best move' based on a given gameboard and search depth. Both the client and server tasks were implemented using PVM [4] and ORBeline [5]. The results indicate that as communication increased between client and server, PVM is faster than ORBeline for the tests and situations presented. As com-putation increased, both platforms performed comparably. Other related areas which are useful to developers designing client-server systems in PVM include Distributed-Object PVM [25] and an RPC facility for PVM [26].