> The design I proposed is to have multiple I/O queues, where to apply the
> elevator, and to choose the queue in function of the task->pid that is
> sumbitting the bh/bio. You'll have to apply an hash to the pid and
> you
Well, if you want *fair* scheduler, as in "fair between users", I
guess you should base it on task->uid.
That should solve your dbench problems, too, as you are very unlikely
to see two tasks working over same data with different uids.
Pavel
-- Worst form of spam? Adding advertisment signatures ala sourceforge.net. What goes next? Inserting advertisment *into* email? - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/