Results 1 to 8 of 8

Thread: Quorum = 1?

  1. #1
    Join Date
    Apr 2008
    Location
    NTSC
    Posts
    536

    Quorum = 1?

    can someone explain what this is (or a link to more info about it)?

    ...and if needed in the current race, how to activate/utilize it?

  2. #2
    Join Date
    Jun 2005
    Location
    Louisa,Virginia
    Posts
    612
    The Quorum is the minimum number of returned results required before points are awarded.If set to 1 then as soon as your results are in you are granted points.

    Tony Kaye, Billy Sherwood, Alan White, Jimmy Haun


    Amd Users StatsJoin World Community Grid


  3. #3
    NeoGen's Avatar
    NeoGen is offline AMD Users Alchemist Moderator
    Site Admin
    Join Date
    Oct 2003
    Location
    North Little Rock, AR (USA)
    Posts
    8,451
    Normally, in most boinc projects, each WU is distributed to two or three people, and only when all of these people return it back, the credits are awarded. This is quorum>1 and is the case when a participant needs "wingmen" (other participants) to crunch and validate the same workunit and only then the credits can be given. Until all results are received and the WU is validated, the credit is pending.

    In quorum = 1, as soon as the result is delivered, it is validated and awarded credit. No need for "wingmen" to validate our workunits. The credits are delivered instantly. This is really good for a race because otherwise people would be stuck with pending credits.

  4. #4
    Join Date
    Apr 2008
    Location
    NTSC
    Posts
    536
    where do i set it?

  5. #5
    Join Date
    Jun 2006
    Location
    vienna, austria
    Posts
    110
    the project admin do this , you can´t change this setting......
    it is wether they "trust" your results to be accurate or not. if they are unsecure they set this to 2 for example, if the two results match(calculated by different machines) you get credits, if not, a third calculation by another one is required to validate the results....
    so it is up to the project admins to do either more research or "securer" research.....
    vel
    mihi deliberandum est....

  6. #6
    NeoGen's Avatar
    NeoGen is offline AMD Users Alchemist Moderator
    Site Admin
    Join Date
    Oct 2003
    Location
    North Little Rock, AR (USA)
    Posts
    8,451
    Yes, validating results is an important part of distributed computing. If under the same conditions two different machines give two different results for the same WU, then one of the machines (or maybe even both) is wrong. So to have accurate results, WUs should be distributed to at least two different machines, and their results compared side-by-side.

    This works great because even in the case that two different machines have a hardware problem, the probability that they have the same problem and produce the exact same error is very very small. So even in the case of both machines being problematic, the results will be different, and so a third (or more) host will be called in for validation.

    And this is the quorum, the minimum number of results needed to validate a WU, as BlackAdder said on the first reply.
    It is not something we can change, the project admins set this as they wish, depending on the level of precision they need for their results. Some projects, of maths for example, may not need high precision in sieving, missing a factor or two may not be the end of the world, and they can always run a double check later, but on other fields of science (biology, physics, etc), projects may need to have accurate results to go further on their investigations.

  7. #7
    Join Date
    Apr 2008
    Location
    NTSC
    Posts
    536
    i'm ASSuming they set a certain (secret) percentage? not just "10 errors and you get Quorum = 2"

  8. #8
    NeoGen's Avatar
    NeoGen is offline AMD Users Alchemist Moderator
    Site Admin
    Join Date
    Oct 2003
    Location
    North Little Rock, AR (USA)
    Posts
    8,451
    I don't think it depends on the number of errors or the users, it's really a global rule set by the project admins when they define what how they want the project.
    If the results precision is critical they must have more results to compare and make sure they are correct.
    If the results precision is not critical (if a few errors won't mind) then they may set quorum = 1.

    Anyway, I'm sure that >99.99% of the machines returns good reliable results. The quorum rule is to prevent the remaining <0.01% of providing bad results that could damage the studies of the project.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •