Long running work unit

Everything about the project RNA World
Nachricht
Autor
Peter Hucker
Mikrocruncher
Mikrocruncher
Beiträge: 30
Registriert: 19.08.2017 13:56

Re: Long running work unit

#877 Ungelesener Beitrag von Peter Hucker » 22.08.2017 21:26

Michael H.W. Weber hat geschrieben:
Peter Hucker hat geschrieben:Also, when I said "why can't the checkpoints be used", what I actually meant was doesn't the server have a copy of where I got up to, so the unit can be handed to the next user half done?
No, the server does in this respect actually not know anything about what the clients are doing. It only receives a trickle-up message how far the individual client is progressing to auto-adjust the tasks deadline on the server-side (run time prediction is actually very difficult for these types of calculations).
.
Odd, I've been given a load of points for the 30% completed task which I aborted when it went back to 0% after a reboot. Does this mean they got useful data or just felt sorry for me?

Jacob Klein
Brain-Bug
Brain-Bug
Beiträge: 564
Registriert: 26.07.2013 15:41

Re: Long running work unit

#878 Ungelesener Beitrag von Jacob Klein » 22.08.2017 21:36

Speed is an 8-core, 16-thread, PC.

I have determined that these RNA World VM Tasks do not hyperthread well with each other at all (verified by looking at "retired instructions" from Intel Performance Counter). So I will only allow 8 to run at the same time, while the remaining threads work on non-VM workloads.

Speed has 6 RNA World tasks currently in progress (5 in BOINC, and 1 that I'm doing locally).
He was already searching for 1 more -- Now he is searching for 2! :)

Peter Hucker
Mikrocruncher
Mikrocruncher
Beiträge: 30
Registriert: 19.08.2017 13:56

Re: Long running work unit

#879 Ungelesener Beitrag von Peter Hucker » 22.08.2017 21:40

Jacob Klein hat geschrieben:Speed is an 8-core, 16-thread, PC.

I have determined that these RNA World VM Tasks do not hyperthread well with each other at all (verified by looking at "retired instructions" from Intel Performance Counter). So I will only allow 8 to run at the same time, while the remaining threads work on non-VM workloads.

Speed has 6 RNA World tasks currently in progress (5 in BOINC, and 1 that I'm doing locally).
He was already searching for 1 more -- Now he is searching for 2! :)
How are you getting so many?

Jacob Klein
Brain-Bug
Brain-Bug
Beiträge: 564
Registriert: 26.07.2013 15:41

Re: Long running work unit

#880 Ungelesener Beitrag von Jacob Klein » 22.08.2017 22:07

Peter Hucker hat geschrieben:How are you getting so many?
For starters, I have set all my other CPU-capable projects to be "backup projects", using 0 Resource Share. Then I set my cache/buffer settings high enough such that I'm always needing work. Then BOINC asks RNA World for work, as often as it automatically can, according to resource backoffs and project backoffs. I also have additional tactics. I can answer more questions privately, if you'd like.

Peter Hucker
Mikrocruncher
Mikrocruncher
Beiträge: 30
Registriert: 19.08.2017 13:56

Re: Long running work unit

#881 Ungelesener Beitrag von Peter Hucker » 22.08.2017 22:13

I see. I've got all my projects with different weights. I guess my machines can't get RNA then settle for something else. Not to worry, I have other projects I like doing too, I was just wondering. One of the four machines is asking every 5 hours, the others aren't asking at all just now.

Jacob Klein
Brain-Bug
Brain-Bug
Beiträge: 564
Registriert: 26.07.2013 15:41

Re: Long running work unit

#882 Ungelesener Beitrag von Jacob Klein » 22.08.2017 22:22

:) I'm a BOINC Alpha tester, attached to all 61 projects. I routinely do work for about 10 of them, and RNA World is one of my top 3 favorite projects. But yeah, I make sure to keep the threads fully loaded, unless I have a PC in "Get 'er done" mode due to upcoming VirtualBox incompatibilities. 2 of my 4 PCs are like that, currently. You might like following the other thread I chime in on, here: viewtopic.php?f=75&t=16160

ChristianB
Admin
Admin
Beiträge: 1920
Registriert: 23.02.2010 22:12

Re: Long running work unit

#883 Ungelesener Beitrag von ChristianB » 22.08.2017 22:27

Peter Hucker hat geschrieben:
Michael H.W. Weber hat geschrieben:
Peter Hucker hat geschrieben:Also, when I said "why can't the checkpoints be used", what I actually meant was doesn't the server have a copy of where I got up to, so the unit can be handed to the next user half done?
No, the server does in this respect actually not know anything about what the clients are doing. It only receives a trickle-up message how far the individual client is progressing to auto-adjust the tasks deadline on the server-side (run time prediction is actually very difficult for these types of calculations).
.
Odd, I've been given a load of points for the 30% completed task which I aborted when it went back to 0% after a reboot. Does this mean they got useful data or just felt sorry for me?
No we don't get the science back but we are rewarding the effort put into trying to tackle a monster by granting partial credit for failed tasks.

Peter Hucker
Mikrocruncher
Mikrocruncher
Beiträge: 30
Registriert: 19.08.2017 13:56

Re: Long running work unit

#884 Ungelesener Beitrag von Peter Hucker » 22.08.2017 23:02

It's a pity you get nothing out of it. Is this a feature missing from BOINC? The inability to upload partial workunits? It must happen often that someone fails to complete a workunit as they go on holiday etc. Or of course in my case a computer error.

Benutzeravatar
Michael H.W. Weber
Vereinsvorstand
Vereinsvorstand
Beiträge: 22419
Registriert: 07.01.2002 01:00
Wohnort: Marpurk
Kontaktdaten:

Re: Long running work unit

#885 Ungelesener Beitrag von Michael H.W. Weber » 22.08.2017 23:28

Peter Hucker hat geschrieben:It's a pity you get nothing out of it. Is this a feature missing from BOINC? The inability to upload partial workunits?
Well, to solve this, there is actually the concept of checkpoints.

Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.

http://signature.statseb.fr I: Kaputte Seite A
http://signature.statseb.fr II: Kaputte Seite B

Bild Bild Bild

Peter Hucker
Mikrocruncher
Mikrocruncher
Beiträge: 30
Registriert: 19.08.2017 13:56

Re: Long running work unit

#886 Ungelesener Beitrag von Peter Hucker » 22.08.2017 23:41

Michael H.W. Weber hat geschrieben:
Peter Hucker hat geschrieben:It's a pity you get nothing out of it. Is this a feature missing from BOINC? The inability to upload partial workunits?
Well, to solve this, there is actually the concept of checkpoints.

Michael.
That's what I thought, so why weren't they used?

Benutzeravatar
Michael H.W. Weber
Vereinsvorstand
Vereinsvorstand
Beiträge: 22419
Registriert: 07.01.2002 01:00
Wohnort: Marpurk
Kontaktdaten:

Re: Long running work unit

#887 Ungelesener Beitrag von Michael H.W. Weber » 23.08.2017 07:21

I think we discussed your specific problem in detail above.

Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.

http://signature.statseb.fr I: Kaputte Seite A
http://signature.statseb.fr II: Kaputte Seite B

Bild Bild Bild

ChristianB
Admin
Admin
Beiträge: 1920
Registriert: 23.02.2010 22:12

Re: Long running work unit

#888 Ungelesener Beitrag von ChristianB » 23.08.2017 20:03

Peter Hucker hat geschrieben:It's a pity you get nothing out of it. Is this a feature missing from BOINC? The inability to upload partial workunits? It must happen often that someone fails to complete a workunit as they go on holiday etc. Or of course in my case a computer error.
BOINC actually does support that but the RNA science app does not. It doesn't do checkpoints so we have to use VirtualBox in order to get checkpointing but we can't upload the VBox checkpoint as it only works on the host it was created on. It would also be too large.

Antworten

Zurück zu „RNA World Discussions (english)“