queue system for > 40 middle class machines ?

queue system for > 40 middle class machines ?

Post by Peter Apian-Bennewi » Sun, 11 May 1997 04:00:00



Hi gurus,
I'm looking for a easy-install, simple queue system to distribute jobs
across 50 machines, running Linux, HPUX and IRIX.
Neither NQS (_all_ programs are setuid root) nor DQS (virtually no docu,
and much config) were found suitable (= ran afer a day's work).
The daemons should figure out as much as possible from the machine
without configs (RAM,swap,paging...) and must allow a script based
configuration (no manuelly-only approach).
Centralized master scheduling is also to be avoided (= 5 points for not
relying on a main node).
I'm pretty close to break down and write my own, please stop me from
re-inventing a round, axle-centralized tranportation device.

If needed, I'll summarize - thanks for all pointers
Peter
---

 Fraunhofer Institute for Solar Energy Systems, D-79100 Freiburg
 http://www.ise.fhg.de/personal_home_pages/apian/Welcome.html

 
 
 

queue system for > 40 middle class machines ?

Post by James Youngma » Fri, 23 May 1997 04:00:00



> I'm looking for a easy-install, simple queue system to distribute jobs
> across 50 machines, running Linux, HPUX and IRIX.
> Neither NQS (_all_ programs are setuid root) nor DQS (virtually no docu,
> and much config) were found suitable (= ran afer a day's work).

Try Condor (http://www.cs.wisc.edu/condor/).

 
 
 

queue system for > 40 middle class machines ?

Post by sa » Wed, 28 May 1997 04:00:00



>> I'm looking for a easy-install, simple queue system to distribute jobs
>> across 50 machines, running Linux, HPUX and IRIX.
>> Neither NQS (_all_ programs are setuid root) nor DQS (virtually no docu,
>> and much config) were found suitable (= ran afer a day's work).

I seem to recall a tcl extension (tcl-dp?) which supports distributed
processing.  Whether this was for parallelizing arbitrary tasks or
just
parallelising tcl I forget ......

hope that helps a bit .....

 
 
 

queue system for > 40 middle class machines ?

Post by Peter Apian-Bennewi » Sat, 31 May 1997 04:00:00


many thanks to all who replied by email; here's the promised summary:

Have you looked at OLTP systems like Tuxedo? Possibly over the top but
they don't have to be complicated if you only use the pieces you need.

If you're willing to use commercial software, there are many good
programs out there.  Look at SysAdmin magazine for the ads.
PowerBroker and Autotrol come to mind.  I don't know if either support
Linux.


http://lars.desy.de/cluster.html


so far so hoopy. Since it didn't sounded very convincing, I started
my own 'poormans-replacement for taskbroker'. Currently in pre-alpha
tests with an urgent video rendering we have to do.

Peter
---

 Fraunhofer Institute for Solar Energy Systems, D-79100 Freiburg
 http://www.ise.fhg.de/personal_home_pages/apian/Welcome.html

 
 
 

1. Lilo "L 40 40 40 40 40 40..."

Greetings,

I have set up a Linux From Scratch system on a 1.2Gb Maxtor hard drive. It
works fine on my two Athlon systems, but when I put the drive in my Pentium
100, the system results in "L 40 40 40 40 40 40..." before showing the Lilo
menu.

My Lilo.conf contains the "lba32" entry which should mean this would work,
but it doesn't.

Any ideas?

Regards,
Jon

2. Install From Parallel Zip Drive

3. LILO saying "L 40 40 40 40 40 40..."

4. at-1.1.tar.Z uploaded to tsx-11.mit.edu

5. Lilo spits "L 40 40 40 40"...

6. OT: Original Designer Returns To Sun

7. My Linux machine doesn't like > 40 users

8. libthread panic... need help!

9. Linux LILO Booting L 40 40 40...

10. AMD 386/40 + ULSI 387/40 crashes intermittantly

11. NewBie Q': Install Linux REDHAT 7 on win98 Machine(MultiBoot) on 40 GB HDD

12. Replace 40 machines with just 1

13. how class A machine(network) to be joined in class D multicasting