Review of The Structure of the "THE"-Multiprogramming system

From: Slavik Krassovsky (viatk_at_windows.microsoft.com)
Date: Wed Jan 07 2004 - 15:20:24 PST

  • Next message: Gail Rahn: "Review of Djikstra's "THE-Multiprogramming" paper"

        In this paper dated 1967 Mr. Dijkstra's descries his approach to
    design, implementation and testing (or verification) of a rather
    advanced for the times operating system THE. The scientific value of THE
    comes from several notions: the notion of sequential processes and
    notion of hierarchy of such processes. Each process provides a service
    of the operating system (processor allocation or memory allocation, work
    with I/O devices, etc.) and possess a level in the hierarchy (with
    processor allocation being level 0 and level 4 being the user programs).
    Only processes of a higher level (bigger number) can request a task from
    a lower level process. Sequential processes are in fact organized as
    cyclic programs with a special wait (or homing) point at which the
    process awaits it's tasks. Essentially the processes are components and
    such componentization in conjunction with the hierarchy organization
    allows first: simplify testing since each component can be tested
    separately and second: theoretically prove that such organization would
    in fact never get into deadlock, infinite loop or "lost task" states.

     

       I really enjoyed reading this paper, although indeed it was proven to
    be time consuming. I command the seriousness of approach of Mr.
    Dijkstra's with regard to an attempt to theoretically assess the
    soundness and correctness of the design, although certainly I found
    myself cautious with regard to his "guarantee of flawlessness" while
    testing is still underway.

     

       Also It was my impression that at the times the paper is written the
    common approach to OS design was monolith complex systems that were
    doing "everything". Such designs might certainly provide some
    performance benefits, however complexity is a concern. Mr. Dijkstra's
    paper clearly demonstrates that along with the complexity reduction,
    component oriented systems bring other values, like ability to apply
    theoretical proofs.

     

     

  • Next message: Gail Rahn: "Review of Djikstra's "THE-Multiprogramming" paper"

    This archive was generated by hypermail 2.1.6 : Wed Jan 07 2004 - 15:20:33 PST