[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-rt
Subject:    Re: [rtl] A realtime project proposal
From:       "maryam moghaddas" <mshsaint () direct ! ca>
Date:       1998-12-20 23:12:33
[Download RAW message or body]

Hi 
Thanx for replying me back and being so concerned.

I will go through what I have done on my project. I hope I'll be clear. If
you have more questions please let me know.

The two schedulers which were originally with the OS were the ones provided
by Mr. Ripoll and Mr. Barabanov. Those two are periodic and event driven.
The scheduler was  preemptive too. What I wanted to test was something
simpler, a non-preemptive bursty EDF scheduler. So it was possible that a
group of tasks arrive all together or in a very short period of time.
Based on this assumption I tried to wrote and tested to schemes for the
scheduler: A task-based scheduler and a batch based scheduler.
A task based scheduler is when the scheduler chooses a task as the best
choice of EDF, it starts execution until completion( since it's non
preemptive). After the task's execution is finished, the scheduler will
check for task arrivals, if yes it will adds to the list of the tasks
waiting for execution(batch) and apply EDF to the list(batch).;Otherwise it
will chose the best EDF from that list
(batch).
In the batch based scheduler, the scheduler will not look for any new task
arrivals until all the tasks in the batch have been executed. Of course the
tasks are chosen for execution with EDF scheme. Mean while the tasks in the
batch are executed it's possible that other tasks have been arrived. So
that batch will start being executed after the previous one is finished.
In both schemes above alot of time is spent on search for the best EDF
choice.
Then I decided to modify the schemes by sorting the batches. So the tasks
will be inserted in on order to the batch upon arrival. In this case all
the batches are sorted based on their deadlines on descending order and
there is no need for search anymore.
This had a great improvement on the scheduling.
The schedulers were tested for set of 500 tasks and 1000 tasks, with
different slack factors. And some neat results were obtained from the
tests.

Happy new year,
I am looking forward to hear more from you...
Maryam



----------
> From: Jose Contreras <jcontrer@if.insa-lyon.fr>
> To: maryam moghaddas <mshsaint@direct.ca>
> Subject: Re: [rtl] A realtime project proposal
> Date: 16 December 1998 01:30
> 
> Hi Maryam,
> 
> could you give more information on the EDF scheduler you developped? A
> report, your memoire, something like that. Meme si le document est en
> francais.
> 
> I do my doctoral program at INSA de Lyon, at the Laboratoire
> d'Ingenierie de l'Informatique Industrielle, and we are working on a
> reflective architecture where scheduling is a key activity.
> 
> I will comment your message with my prof, next friday.
> 
> Keep in touch!
> 
> Jose
--- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail majordomo@rtlinux.cs.nmt.edu OR
echo "unsubscribe rtl <Your_email>" | mail majordomo@rtlinux.cs.nmt.edu
----
For more information on Real-Time Linux see:
http://www.rtlinux.org/~rtlinux/

[prev in list] [next in list] [prev in thread] [next in thread]