Review: Programming Semantics for Multiprogrammed Computations.

From: Sellakumaran Kanagarathnam (sellak_at_windows.microsoft.com)
Date: Mon Jan 12 2004 - 12:47:07 PST

  • Next message: Manish Mittal: "Review:Programming Semantics for Multiprogrammed Computations"

    The authors begin with citing three then current different examples of
    multiprogrammed systems and it was very interesting to note that an
    airlines reservation system was an example.
    The authors start with five properties of multiprogrammed computer
    systems, explain the concepts and terminology that they are using,
    supervisor (core of basic computer system functions), primitive
    operations used for parallel processing. They then go on to explain what
    they call as inferior spheres of protection, protected entry points and
    end with a detailed explanation of the directories and naming system.
    The five properties of MCS are: 1) concurrent operations for multiple
    users 2) sharing pools of resources with problem of efficiently using
    resources taken out from the users 3) demand for computing resources
    vary and hence the amount of physical resources required is governed by
    average demand rather than peak demand 4) common information shared by
    multiple computations 5) MCS should be a modular multiprocessor system
    so that modules can be added later on.
    The authors go on to define the concepts and terminology. It is nice to
    see security has been given due consideration at various levels
    (different spheres of protection - c-list and so on). In this paper
    also, a segment refers to both memory as well as retained objects
    (files). A computation seems to correspond to a process in modern world
    and a process seems to correspond to a thread.
    A principal seems to represent individual users, a group and/or a
    specific function.
    The supervisor forms the core basic block of the MCS. Some of the Meta
    instructions like fork and quit sound familiar but instructions like
    join do not.
    In discussing the inferior spheres of protection, it is very interesting
    to note the exception handling facilities provided.
    Directory naming: retained objects. The constraint that retained
    objects' name cannot be changed reflects poor design and the second
    reason given is a no reason.
    The initial solution suggested for dealing with ambiguous names (an
    awkward restriction being imposed), all these tend to imply that this
    paper is one of the initial attempts to define characteristics and Meta
    language for MCS.
    Even though the paper refers to three practical systems and often refers
    to MAC system, the ideas represented seem to be more of conceptual
    nature and no referent to an actual system based on the concepts is
    made.
    The paper attempts to define a list of Meta instructions that can be
    used in an MCS and while doing so it explains some of the concepts and
    terminology used.


  • Next message: Manish Mittal: "Review:Programming Semantics for Multiprogrammed Computations"

    This archive was generated by hypermail 2.1.6 : Mon Jan 12 2004 - 12:46:59 PST