Informatikk:
Research and Teaching in Norway
A Critical Evaluation

NAVF: The Council for Natural Science Research 1992, ISBN 82-7216-813-0

Page 34-35. Copied with the authorization of NAVF, Rune Dyre, by Petter Reinholdtsen - petterr@stud.cs.uit.no

Distributed Operation Systems and Distributed Computations

Permanent scientific positions

2 Associate Professors: Anshus, Hartvigsen
3 Assistant Professors: Fallmyr, Johansen, Larsen

Students

2 Doctoral students (1 is assisting professor and 1 is associated professor)
23 Graduate students:
- 14 Cand. scient. (4 of these have principal supervisor outside the department)
- 9 Siv.ing. (1 of these has principal supervisor outside the department)

Theses last 5 years (1988-may 1992)

13
- 6 Cand.scient. theses and
- 7 Siv.ing. theses

Research activity

The research of this group concerns distributed operating systems, both for networks and for parallel processors. The three research components are:

MacroScope Project,
Monitoring Project, and
StormCast Project.

The MacroScope Project analyzes abstractions to see how they can support monitoring and control of large distributed systems. The group is building a new kernel of an operating system that provides a shared-memory model using messages-based communication. It can reduce the information transmitted using a new abstraction for state vectors. The group is building a file system and a scheduler to support the shared-memory architecture as well.

The Monitoring Project wants to evaluate the performance of monitors in distributed systems, specifically by the creation of a distributed monitor architecture. Their project also develops languages to specify states and events to be monitored.

The StormCast Project includes distributed applications and expert systems. The application currently investigated involves many sensing devices for a Cray supercomputer that does weather forecasting. The weather predicted by the Cray is broadcast twice a day to all ships at sea. As the weather predicted uses a large grid over a region, the weather at the ship may be very different from the predicted weather for a region. The expert system takes the predicted weather, the data from the sensors on that ship, and the information transmitted from other ships to adjust the local weather prediction. The StormCast application will be used to evaluate the distributed monitors mentioned above.

Observations

The group is young, enthusiastic and is attacking an important task. This youth likely explains the lack of appreciation for a sharp focus to their research. Having an application to drive the system work is an excellent idea. Each member for the group knows what the others are doing, and as a result they are trying to use each other's concepts.

The group has brought in an impressing list of visitors in distributed operating systems; the primary reason seems to be that they organized four workshops concerning distributed systems in the last four years in or near Tromsų.

Perhaps as a result of the workshops, the group seems to be well informed about the work being done internationally in this area. However, they are building systems similar to systems being built by others, and so there is the danger that their work may not lead to publishable results or to doctoral dissertations.

They have a nice model of developing software to be used in the field. Prototype software is developed at the university and then given to the company responsible for the commercial version of the software that will run on the ships. This allows the group to develop ideas and to see them used without having the responsibility of software maintenance.

Recommendations

"Storm Cast" is a very ambitious project, but the group is working on an interesting approach. There is concern that the group did not have well defined scientific goals at the start of the project. The group should identify those goals soon.

There is also concern that in the rush to finish an ambitious project research objectives will be abandoned, or that results achieved will not be published.

Publishing has to be an important concern for this project. It will help to get Ph.D.s for faculty members, which is also a high priority item recommended to the group.