Review: Waldspurger & Weihl, Lottery Scheduling

From: Steve Arnold (steve.arnold4_at_verizon.net)
Date: Tue Jan 27 2004 - 21:27:40 PST

  • Next message: ahemavathy: "Lottery Scheduling Review"

    This relatively short paper describes basically an algorithm for scheduling
    within an operating system. The authors leave it fairly open-ended for what
    resources are scheduled. The whole thing is based on a lotter metaphor,
    where each task contending for resources holds lottery tickets with a preset
    value. A random number generator then "picks the winner." The winner gets a
    certain number of quantums on the resource. It is possible to have a number
    of currencies on the system. Each currency runs its own lottery, as I
    understand it.
     
    In their testing, they show that the lottery is quite fair, despite the
    random aspect of it. It also has about the same overhead (slightly less) as
    competing scheduling systems. The authors really do a lot of tests to
    basically demonstrate the design of the system (which could have as easily
    been done mathematically as experimentally).
     
    This was one of the more straight-forward papers we have read to date. It
    basically shows that a random algorithm, in most cases, is just as good as
    much more complicated systems. They do complain that priorities are rarely
    used properly in other systems; I wonder if it will be the same with this
    one. I would like to know where they took this research and if it has been
    implemented in any production systems.
     


  • Next message: ahemavathy: "Lottery Scheduling Review"

    This archive was generated by hypermail 2.1.6 : Tue Jan 27 2004 - 21:27:35 PST