Deadline too short

Everything about the project RNA World
Nachricht
Autor
marsinph
PDA-Benutzer
PDA-Benutzer
Beiträge: 31
Registriert: 05.04.2018 09:08

Deadline too short

#1 Ungelesener Beitrag von marsinph » 12.08.2019 13:53

Hello,
A wonder : today I got one WU !
http://www.rnaworld.de/rnaworld/result. ... d=14954578

Nice/, but
Estimated computing time 83 days on my I7-3930K OC 4.2Ghz (host 74429 )
After six hours : 0.23% done ! So it would be about 100 days.
And it runs, I see the time of checkpoint each 30 minutes.
Also change in slot directory.

The problem : the deadline is on 01 sep 2019 !?

What now ?

Benutzeravatar
compalex
Partikel-Strecker
Partikel-Strecker
Beiträge: 999
Registriert: 13.12.2016 17:54
Wohnort: Hockenheim

Re: Deadline too short

#2 Ungelesener Beitrag von compalex » 12.08.2019 17:37

you have to sprint right into the shop and buy a Ryzen 3900X! :roll2: :lol:
Gruß memorix//Alex

Hardware on DC:
#1 Ryzen 3700X, 32GB, GTX 1080
#2 Ryzen 3700X, 32GB, GTX 1080
#3 Ryzen 3600, 16GB, RTX 2070
#4 Ryzen 3600, 32GB, 5*GTX 1070Ti
Bild
Bild

Benutzeravatar
yoyo
Vereinsvorstand
Vereinsvorstand
Beiträge: 7788
Registriert: 17.12.2002 14:09
Wohnort: Berlin
Kontaktdaten:

Re: Deadline too short

#3 Ungelesener Beitrag von yoyo » 12.08.2019 19:28

The app sends trickle messages to the server and beginning 14 days before the deadline it shifts the deadline always 14 days ahead. So if your wu is running at least once in 2 weeks the deadline is always moved ahead. But you see the changed deadline only on the server web page.This is not synchronized to your BOINC client.
HILF mit im Rechenkraft-WiKi, dies gibts zu tun.
Wiki - FAQ - Verein - Chat

Bild Bild

Benutzeravatar
Michael H.W. Weber
Vereinsvorstand
Vereinsvorstand
Beiträge: 20369
Registriert: 07.01.2002 01:00
Wohnort: Marpurk
Kontaktdaten:

Re: Deadline too short

#4 Ungelesener Beitrag von Michael H.W. Weber » 13.08.2019 10:28

...and keep in mind that the estimated run times are usually to be doubled in case of AMD systems.

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

Benutzeravatar
gemini8
Vereinsmitglied
Vereinsmitglied
Beiträge: 3271
Registriert: 31.05.2011 10:30
Wohnort: Hannover

Re: Deadline too short

#5 Ungelesener Beitrag von gemini8 » 13.08.2019 13:47

Well, we don't know yet about the new ones.
Gruß, Jens
- - - - - -
Lowend-User und Teilzeitcruncher

Bild Bild
Bild

marsinph
PDA-Benutzer
PDA-Benutzer
Beiträge: 31
Registriert: 05.04.2018 09:08

Re: Deadline too short

#6 Ungelesener Beitrag von marsinph » 13.08.2019 19:09

Hello Yoto, Michael,
Thank you for explanation. I hope about report (additional time) of deadline will be real.

But why to produce such "monster" of WU ?
If you look the WU, about 60% of users cancel it and 40 "error". Only less than 1% could be finished
At this time already 140 resent !
The WU I have , was made in 2013. Six years ago and nobody finished it.
http://www.rnaworld.de/rnaworld/workuni ... id=6330825
And additional, it also need a wingman to be validated.

Only one have let it fully run 19,053,214 sec ( about seven months) to get a "inconclusive" !!!
Nice job for the research. Of course, we get credit. But scientists : nothing

Honnestly, you think this WU will ever be fully finished ? In ten years, perhaps ?
This project is also more than six years in "beta" phasee. Like main page write.
Also announce very long WU above 1.000 hours (41 days), ot seven months

Then server status reports 57 tasks in progress. But 6.371 task waiting deletion. But server is already months ago "disabled"
Look also amount of recent active users : 23. Recent acrive hosts : 20.
Look your own nice graphics on server status. Nothing


Why , you produce such WU ? I thik scientists need results and fast. They not need years to wait for nothing.
It is the same on Yoyo@home, Yafu : all "monsters" unstable, who not follow the limitations of users also not limitations of Boinc settings.

I will let turn my WU

It was a long post to try to show admin, that, if their scientist wants results, it is perhaps not the best way to do.

Best regards

Benutzeravatar
compalex
Partikel-Strecker
Partikel-Strecker
Beiträge: 999
Registriert: 13.12.2016 17:54
Wohnort: Hockenheim

Re: Deadline too short

#7 Ungelesener Beitrag von compalex » 13.08.2019 19:49

hi marsinph, if you do not have the patience for RNA, then you should switch to PrimeGrid. There are long and short WU runtime projects and everyone is free for which he decides. You have been with RNA for years and know the procedure exactly, because you do not need to stink in every forum for every project. :roll:
Best regards
Gruß memorix//Alex

Hardware on DC:
#1 Ryzen 3700X, 32GB, GTX 1080
#2 Ryzen 3700X, 32GB, GTX 1080
#3 Ryzen 3600, 16GB, RTX 2070
#4 Ryzen 3600, 32GB, 5*GTX 1070Ti
Bild
Bild

Stiwi
Projekt-Fetischist
Projekt-Fetischist
Beiträge: 663
Registriert: 20.05.2012 21:11

Re: Deadline too short

#8 Ungelesener Beitrag von Stiwi » 13.08.2019 19:51

If a calculation took so long and can't be splitted you can not do some magic and the task will finish a few hours.

From the mainpage:
The runtime of the cmsearch VM apps are very long (>1000 hours) and deadlines are automatically extended but only visible on the website!
If you don't accept this you shouldn't run this project.
Bild
Bild

Jacob Klein
Oberfalter
Oberfalter
Beiträge: 481
Registriert: 26.07.2013 15:41

Re: Deadline too short

#9 Ungelesener Beitrag von Jacob Klein » 13.08.2019 20:26

I did an estimation using my calculator in another thread, based on the work unit's previous completion.
I think it will take your PC about 220-280 days of CPU Time to complete this task.
But it could take up to 500 days.

Ignore the BOINC values of "Progress %" and "Remaining (estimated)". They are useless. The task will eventually get to 98.765%, then run there for several days. It will eventually complete. You must be gentle and patient. The overclock will help a little bit, maybe 10-20% quicker.

Regards,
Jacob Klein

Code: Alles auswählen

Threads: You All, Wingman All:		23,917,115	276d 19:38:35
Threads: You All, Wingman Half:		40,558,411	469d 10:13:31
Threads: You All, Wingman Single:	42,743,056	494d 17:04:16
Threads: You Half, Wingman All:		13,394,658	155d 0:44:18
Threads: You Half, Wingman Half:	22,714,531	262d 21:35:31
Threads: You Half, Wingman Single:	23,938,030	277d 1:27:10
Threads: You Single, Wingman All:	12,335,545	142d 18:32:25
Threads: You Single, Wingman Half:	20,918,498	242d 2:41:38
Threads: You Single, Wingman Single:	22,045,256	255d 3:40:56
BOINC Measured FLOPS (million ops/sec):	17,941,893	207d 15:51:33

Benutzeravatar
Michael H.W. Weber
Vereinsvorstand
Vereinsvorstand
Beiträge: 20369
Registriert: 07.01.2002 01:00
Wohnort: Marpurk
Kontaktdaten:

Re: Deadline too short

#10 Ungelesener Beitrag von Michael H.W. Weber » 14.08.2019 09:30

marsinph hat geschrieben:
13.08.2019 19:09
This project is also more than six years in "beta" phasee. Like main page write.

Then server status reports 57 tasks in progress. But 6.371 task waiting deletion. But server is already months ago "disabled"
Look also amount of recent active users : 23. Recent acrive hosts : 20.
Look your own nice graphics on server status. Nothing

Why , you produce such WU ? I thik scientists need results and fast. They not need years to wait for nothing.
It was a long post to try to show admin, that, if their scientist wants results, it is perhaps not the best way to do.
Our publication list so far: https://www.rechenkraft.net/wiki/Verein ... blications

I would like to add that beside these, we have presented a couple of very interesting research result posters at various highly reputated international conferences and institutes (including MIT/Cambridge/USA).

On top of that our experimental work (which our contributers made possible!) allowed us to give detailed feedback to the original developers of the underlying software to improve their software. They had no idea about excessive run times, they did not know that some cmcalibrate tasks may take >128 GB of RAM (and this as of the year 2009!) and they did not care about implementing checkpointing because they are mathematicians having virtually infinite compute ressources available as HPC at any given time. And, because they are kind people, most of your suggestions were implemented in the updated software versions making it feasable for non-DC projects to use it for the first time to carry out stand-alone desktop PC bioinformatics calculations to advance RNA research. And finally, we compiled the first Windows (and even ARM) clients for the entire set of their software developments. 8)

We also experienced better funded work groups publishing results in high-class journals earlier than us - results which we found earlier and in a much broader manner. This is quite sad but that's life, they just get paid for dong nothing but this and consequently can put more work power into this interesting field of science.

So, trust me: We know what we do and why. And we are NOT a team of well-equipped scientists who get paid for this work - we play with this in our spare time. As a result, things take much, much longer as desired. And for that, I think we are doing quite well.

From reading your postings I detect some sort of frustrations. I interpret these as a desire to contribute to our project and for that I would like to thank you. :good: My suggestion is to just wait until we can roll out future tasks which will again require less computational efforts than the current work units.

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

marsinph
PDA-Benutzer
PDA-Benutzer
Beiträge: 31
Registriert: 05.04.2018 09:08

Re: Deadline too short

#11 Ungelesener Beitrag von marsinph » 14.08.2019 19:46

Hello Michael.
I really, honnestly thank you very much for your long answer.
I not discuss about scientifical results. I not discuss about the interest of RNA.

The only, I repeat the only is about the WU it self. Nothing else. Once again the WU itself.

I understand it is not easy for you to make it, to program (or split) the WU.
As very youg retired, i have much more time to analyse. I suppose you see who I m, where I crunch, also inside team.
Once again, WU 6330825 , nobody have finish it (since 2013). This WU ill never finish !
Not forget, we also change of computer, hard disk,.... . One user have let run 19,000,000 sec (task 14953392, host 29376) seven month 24/24 to get "inconclusive"
It is why, there are very few active users.
Once again, I have one WU, and I let run it (WU 74429, task 14954578) with hope it can fully finish. host 24/24, 7/7

So consider the creation of WU....

Best regards

Philippe

marsinph
PDA-Benutzer
PDA-Benutzer
Beiträge: 31
Registriert: 05.04.2018 09:08

Re: Deadline too short

#12 Ungelesener Beitrag von marsinph » 14.08.2019 20:13

gemini8 hat geschrieben:
13.08.2019 13:47
Well, we don't know yet about the new ones.
Gemini8, i think it would be better ou return results instead of posting without any knownledge !
2993 post and no any results, you not exist on BoincStats, but the most knownledge here on RNA !
https://www.boincstats.com/stats/-1/use ... rojectList

I have no patience, perhaps not enough. You are fully right.
But you do nothing !!!
Today only ione GPUGrid returned.
About 10 WU on LHC
On TN : 2 WU
On Yoyo also 2 WU
And on NFS 10 WU
On PrimeGrid, you produce in one mont what I produce in one day.
On Collatz, you produce in one mont what I do in a few hours.....
And so one... und zu weiter....
Let me be clear, it is not about computing power, only about knownledge of project. Nothing else.

But here nothing, and you posts about two thousand recommandations !?!?
Start to crunch on RNA, then analyse all, then, we ill discuss !
Schone abend (please not answer in German, I am not able. FR, NL, EN, OK

Antworten

Zurück zu „RNA World Discussions (english)“