[RC5] Interactive jobs detectors

Francis Muguet muguet at enstay.ensta.fr
Fri Nov 20 21:25:00 EST 1998


Att: rc5 at lists.distributed.net 

 Dear distibuted netters, 

 I seek your help and advice on distributed computing constraints.

 A colleague of mine is running rc5 on Linux PCs dedicated
 primarily to teaching.
 I am using also these PCs for my research.
 Research jobs should have of course a high priority over
 rc5, mine are taking at least 40Mb ( 128Mb PC ), and
 should affect in the least possible way 
  student interactive jobs.
 However even with rc5 is running at nice+20, it takes a 20%
 CPU load over my own research jobs which are running at nice+7
  ( before at nice+10, this was slowing down too much my programs ).
  But even this nice+7 value is too high if I want not to affect
  student interactive jobs.

 I would hate to ask my colleague to stop rc5 on our Linux PCs,
 since I think that using idle CPU time is a good idea.
 So, I wonder if there is some neat solution to these constraints.

 I remenber having read a discussion about the use of idle
 Unix computers to prove a graph theory theorem
  ( 4-color theorem I think ). The author 
 was exposing that the use of "nice" was not adequate and
 was using yet another method I forgot.

 The SETI project jobs are going to run as "SreenSavers" as I read
 on their WWW site, and this should constitute a better solution.
 In fact, I just registered to SETIatHome.

 Is there any Unix software which can help us in the current
 situation :
   1/ run only after no interactive activity has been detected
   for a minute or so.
   2/ enable to have different level of priority with these
   "screen savers" jobs or better have 
   "screen savers" of different hierachies..
   research jobs running after no interactive being active for a while.  
   distributed computing rc5 jobs running only 
   after no research or interactive jobs being active for a while.
 
 I have just sent a mail to r5help,
 the quasi-instantaneous  (I was stunned !) answer
 was :

> jobs.  As for further automation, I can not help.  You should e-mail
> rc5 at lists.distributed.net and ask there; I'm quite sure there are some
> people on the list who have shell scripts/etc that do load
> detection/etc.

  Plase answer directly to me, since I do not belong to the list.

  Many thanks

  Francis



===================================================================
Francis F. Muguet Ph.D
Chercheur-Enseignant   /  Research Scientist and Lecturer
Laboratoire de Mathematiques Appliquees
ENSTA ( Ecole Nationale Superieure de Techniques Avancees ) 
32 Boulevard Victor
75739 PARIS Cedex
FRANCE

e-mail: muguet at ensta.fr  muguet at enstay.ensta.fr  
Web :  http://yvette.ensta.fr/~muguet
Phones : (33) 01.45.52.60.19 - (33) 01.69.31.98.76 - (33) 06.82.30.26.70
Fax : (33) 01.45.52.52.82
Office : room 3075 
------------------------------------------------------------------
also member of : 
ASCI ( Applications Scientifiques du Calcul Intensif )
UPR-CNRS 9049
Bat. 506  Campus d'Orsay
91403  ORSAY
FRANCE 
e-mail: muguet at asci.fr     Phone: (33) 01.69.35.84.16 
Web :  http://www.asci.fr/~muguet      (fastest acess)
------------------------------------------------------------------
also member of :
GDR-CNRS  1017, European PICS 323
=================================================================== 

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list