A Distributed System with a Centralized Organization

One of the most important potential benefits of workstation clusters (client/server computing systems) is resource sharing. By interconnecting a number of workstations using a suitable network, a large number of hardware and software resources can be made available to users. In comparison with centralized computer systems (mainframes) and shared memory multiprocessors systems, the availability of such resources can be achieved at a fraction of the costs entailed by mainframe solutions, since workstations allow the use of standard components. Another advantage of workstation clusters is their scalability and flexibility towards an upgrade or extension with some other components. The concept of a mainframe system is sharing centralized resources among all users by switching between users’ processes (time sharing). The main advantage of such systems is that they are usually easy to administrate. In contrast to this concept, we will introduce a distributed system that does not only offer time sharing, but also sharing of computing nodes (workstations) according to the users’ application requirements (workstation sharing). Nevertheless, the organization of that system remains centralized in order to reduce the expense for management and maintenance. In this paper, we will explain the architecture and mechanism of a transparent system that consists of some Unix workstations running a network file system (NFS).