Next: Computer Networks Laboratory Up: Department of Computer Previous: Distributed Parallel Processing

Operating Systems Laboratory


/ Runhe Huang / Assistant Professor

The Operating Systems Laboratory is participating in both top-down education and state-of-the-art research.

In education, the final phase of development of a workstation-based laboratory has been completed. The laboratory has a special layout for programmers to share 19 Sun-based workstations and 4 X terminals, connected to the laboratory file server via a network. In addition, a multi-media Macintosh has a scanner, video camera, video recorder, and television monitor.

An office for students has been built next to the laboratory for lab assistants, lab members, and visiting students. The lab assistants are exceptional students who help with lab software and exercises; the lab members are students who have made significant progress on their projects; and the visiting students are from other laboratories who are doing temporary work in the Operating Systems Lab.

In addition to concurrent programming, the laboratory has developed special simulation tools for students to learn the practical and theoretical aspects of queueing systems. The software is written in a new high-level script language (Tcl/Tk) that allows the rapid-prototyping of software, particularly educational software. The software has been used in the Performance Evaluation course and another tool has been used in the Database course.

In research, the major themes of the research are stability, adaptation, and group formation under the constraint of delayed information. The general goal of the research in the laboratory is to establish principles that improve the performance of highly distributed systems. The systems require adaptability, reliability, and expandability; the solutions are characterized by self-organizing and learning behaviors. The agents, or controllers, in the system make autonomous decisions and are expected to maintain reasonable performance, even as the system operates under aged or misleading information or evolves into unforeseen environments.

Some of the topics covered are the

  1. value of localized decision making in decentralized control;
  2. sharing of global resources among autonomous agents;
  3. use of randomization in load balancing queueing systems;
and some of the specific topics on stability are the
  1. stability of dynamic groups in distributed systems;
  2. stability of computational ecosystems with queues and learning automata;
  3. stability of goal-directed versus response-directed algorithms;
  4. stability of adaptive methods in hierarchical game structures;
  5. stability of a prisoner's dilemma with delayed information;
  6. stability of database transaction processing.



Next: Computer Networks Laboratory Up: Department of Computer Previous: Distributed Parallel Processing


www@u-aizu.ac.jp
November 2000