ACM Computing Surveys (CSUR) - The MIT Press scientific computation series
Communications of the ACM
The distributed operating system of the SMD project
Software—Practice & Experience
Performance of the world's fastest distributed operating system
ACM SIGOPS Operating Systems Review
Measurements of distributed operating systems
Proceedings of the Twenty-First Annual Hawaii International Conference on Software Track
Asynchronous communication on Occam
ACM SIGPLAN Notices
Simulation & design of message routing systems for network of transputers
ANSS '89 Proceedings of the 22nd annual symposium on Simulation
Supporting the development of multi-agent interactions via roles
AOSE'05 Proceedings of the 6th international conference on Agent-Oriented Software Engineering
| Hi-index | 0.00 |
Distributed computing (variably - and confusedly - called parallel computing or concurrent computing) has reduced the quantum of discretization of computing power. No longer do computers cleanly can into the clearly demarkated categories of Microcomputer, Minicomputer, Mainframe and Supercomputer. By the simple expedient of putting together a variable number of processors (of variable size/power), we now have computers falling almost continuously throughout the range from micros to supercomputers. New terms have been coined for intermediate ranges, like SuperMinis and MiniSupers. Distributed computing is gradually being accepted as the dominant high-performance computing paradigm of the future. Once again, however, hardware seems to have left software behind. While it is fairly straightforward to build a distributed computer, developing efficient software for it appears to be a different ballgame altogether. Software technology for centralized operating systems is stable and reliable, but tight coupling (shared memory) puts an inherent limit on the degree of parallelism, and hence on the performance of parallel computers. It is, thus, very important to establish a firm understanding of the concept of distributed computing and distributed operating systems (DOS).Distributed operating systems (where "distributed" implies distribution of functionality and control) are generally said to be of two types: DOS/RPC (Remote Procedure Call) and DOS/MP (Message Passing). While variants of these two abound, no strong third alternative is seen today. For the loosely-coupled architecture, RPC at a greater depth reduces to message passing. It is thus of fundamental importance to fully explore and exploit the paradigm of message passing. This article attempts to give a brief but encompassing overview of distributed message passing operating systems.