Interoperable object models for large scale distributed systems

Most legacy client/server systems have been written with data and functionality bound in a non-ideal way resulting the user having to manipulate the input data and interpret the results, usually using 4GLs. Ideal client/server applications would contain more appropriate bindings of functionality and data into objects that can be made freely available over networks. One possible approach to these problems is to develop client/server solutions that are based on interoperable object models. These are designed to hide the complexity of heterogeneous distributed systems, and address the desire of the user to have seamless integration of applications and data wherever they may be physically located. The two tier client/server model generally uses a DBMS that provides the server and manages the client network connection. Both the client and the server co-operate in executing the application, but the client contains all of the application logic. The server contains the database management system which hides the complex functions that are needed to manage the data. The data resides on the server and the application logic resides on the client. The distributed communications are therefore relatively easy to manage. The weaknesses of the two tiered model is that the processing balance is constrained by the separation of client logic from the server data. The two tier model can be characterised as islands of data isolated from each other. The three tier model introduces an additional element of 'middleware' between the client and the server. The middleware runs on all machines that host a client or server. The task of the middleware is to provide services to clients or servers to enable efficient delivery of operation requests and return of results. Middleware may typically provide other features. Four types of middleware have been identified by Orfali. These are transport stacks such as TCP/IP and NetBIOS; network operating systems such as the distributed computing environment; distributed system management such as the Distributed Management Environment; and service specific middleware. The characteristics of service specific middleware are determined by the server requirements. This paper surveys different approaches to extending the client/server model towards an interoperable object model The models surveyed are the X-window client/server model, the ISO Open Distributed Processing model, the OMG Common Object Request Broker Architecture model and the IBM System Object Model. Finally, an evaluation of an interoperable object model selected for experimental implementation of a simple distributed client/server application is given.