Operating Systems of the 1990s and Beyond

Motivation

Architectural and hardware advances in computing systems design are occurring at an ever quickening rate, but it is not clear that the operating systems that make these new systems useful are keeping pace. Indeed, it is the operating system that masters the complexity of the ever more complex computing devices being built to make them useful tools. In the past, the existence of a variety of operating systems has made the difference between an interesting architecture and a useful computing environment.

Today, as more and more complex computational structures are emerging, and new and more powerful communication technologies are becoming available, we are faced with the need to develop new generations of operating systems to harness their power. A few of the new challenges to face the operating system implementor include, but are not restricted to:

• Unreliable communications,
• Fault tolerance,
• Issues of size and scalability,
• Integration of heterogeneous systems,
• Supporting advanced applications, such as multimedia,
• Protection and security issues in faulty/untrustworthy distributed systems,
• Coping with existing systems in networked environments.

What form should future operating systems take to address these and numerous other complex problems? In which area is further research indicated? Is the current generation of operating systems a valid platform for the operating systems of the next decade and beyond, or should we be designing a whole new generation of operating systems from the bottom up? What type of architectural support for operating systems and communications hardware should be built into the next generation of computers?

These are general questions that our workshop attempted to address. It is our belief that now is the time to address these questions in a manner that will have some chance of producing useful results in the form of guidelines for future operating systems design and development. Three compelling circumstances lead us to believe that this is the time to act in terms of plotting a rational course for operating systems of the 1990s and beyond:

• The commercially available operating systems of today, including UNIX, have been designed, or based on designs, that are twenty years old, with time-sharing on a single node machine being the foundation of the design. Networking and distributed computing facilities were "add-ons" and not part of the basic system design. Today's new computing architectures are moving in different directions, unfortunately, with operating systems that have been pieced together to do the job.
• UNIX has formed the foundation for a de facto international standard for operating systems. Is UNIX the proper foundation for future efforts? Is its basic structure appropriate for computing equipment currently coming onto the market, or more importantly for new machines that will be unveiled in the near future?
• We are now at a crossroad characterized by a change in technology from traditional single-node systems to networks of computers, distributed machines and massively parallel systems. It is not clear that current operating systems technology is poised to take advantage of these new machines when they do become available.

For all of the above reasons, an intemational workshop entitled "Operating Systems of the 1990s And Beyond: Where Do We Go From Here" was organized and held at the Intemational Conference and Research Center for Computer Science (IBFI) at Dagstuhl Castle in the Federal Republic of Germany. The Dverriding motivation behind the workshop was to provide an opportunity for a relatively small number of leading researchers in operating systems from both universities and industry to meet and discuss current problems and future directions.

The workshop was structured into several working sessions and one final session which was planned as a workshop summary. During each session, a white paper was presented followed by several position papers, all of which were mingled with open discussion. The complete material including the text of all white and position papers as well as a condensation of the discussion is contained in the workshop proceedings (A. Karshmer, J. Nehmer (Eds.), "Operating Systems of the 90s and Beyond", Proceedings, LNCS 563, Springer-Verlag). In the following, abstracts for each white paper presentation are given.