Re: Fwd: Programming bounty



On Wed, Jul 23, 2008 at 10:44:13PM -0400, Dan Heily wrote:
> By modeling the network of tasks, the order is automatically determined.
>
> Forward pass:  "start" all tasks that have no predecessor (assuming resource
> availability). 

It's the above assumption of resource availability that I'm talking
about. If you have two tasks that can both start, but require 100% of
the same person/resource, then you need to make a decision about task
ordering. What you decide can drastically change the makespan of the
schedule. It is the algorithm that makes these decisions that is not
trivial.

In fact these algorithms are about more than just making a decision from
time to time. They are usually a complete solution (using things like
genetic algorithms) that give you a complete ordering of tasks at the
end. I believe this makes it difficult to do things like task splitting
if it isn't already part of the algorithm.

I hope I'm not glossing over the rest of what you wrote, but I think I
understand and agree with that part of it.

> The epitome of my simulations was a machine shop simulation that would roll
> back lower priority tasks that had allocated a resource that was later
> needed by a higher priority task.  The process flows included forks/joins
> whereby a process could split and later synchronize.

This is what I thought Kurt Maute's priority scheduling would be like.
The advantage of this approach, besides the fact that tasks can easily
be interrupted, is that the user can manually tweak the schedule by
changing priorities of tasks. I still think this is a simple, yet useful
way of scheduling.

Regards,
Maurice.

-- 
Maurice van der Pot

Gentoo Linux Developer   griffon26 gentoo org    http://www.gentoo.org
Gnome Planner Developer  griffon26 kfk4ever com  http://live.gnome.org/Planner

Attachment: pgp3dVitGXYEu.pgp
Description: PGP signature



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]