Distributed Objects in Nó//

This paper presents the object access control and distribution strategies for the Nó// (reads parallel node) project, which aims to develop a high performance, low cost, multicomputer and a concurrent programming environment for it (Corso, 1993). Sharing resources is the main goal of multiprogramming environments (Andrews, 1990). Therefore, operating system must provide control mechanisms to make object access safety and efficient. Although there are well-known mechanisms for single memory systems, they are inadequate for distributed memory systems like Nó//. Nó// operating system is projected to have an automatic and transparent computational load distribution. In this context, process and other operating system objects, like semaphores, mail boxes and still memory segments can freely migrate from a processor to another one. In order to allow that, without using inadequate broadcasting mechanisms (Tanembaum, 1992), Nó// operating system propose the indirect capability abstraction, which is presented in this paper. The text first gives an overview of the whole project and shows how computing power can be enhanced by dynamic load-balancing system. Then, indirect capability concept is depicted to demonstrate how it allows security and transparent migration of processes and other operating system objects, improving system performance. In this context, it is discussed how Nó//’s process concept (Fröhlich et al, 1996) can be extended to allow distributed threads. Finally, it is presented a perspective analysis of distributed system evolution highlighting reliability, flexibility and efficiency.