Advanced search

Message boards : Server and website : Current Server Status

Author Message
Bedrich Hajek
Send message
Joined: 28 Mar 09
Posts: 467
Credit: 8,190,146,966
RAC: 10,558,941
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 46054 - Posted: 4 Jan 2017 | 0:33:47 UTC

The current server status shows over 11,000 long runs, over 2,000 are the GERARD_MO tasks, while over 9,000 are other tasks. Yet, the last few days, I have been getting mostly GERARD_MO tasks, which are slow and have low GPU usage, and a refuse to crunch them. When are the other tasks going to be released?

And a more important question is: Why can we have the right to chose which task type, we want to crunch and which we don't want to crunch?
See link:

http://www.gpugrid.net/forum_thread.php?id=4366#44307


And lastly, if you are offended by my action and statement, I don't care!





Erich56
Send message
Joined: 1 Jan 15
Posts: 1090
Credit: 6,603,906,926
RAC: 21,893,126
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 46058 - Posted: 4 Jan 2017 | 11:32:42 UTC

well, this leads to the interesting question whether there is a system by which the tasks are distributed, or whether it's totally up to coincidence which task a cruncher receives.

I am asking this because I noticed the following:
although in the recent past, the total number of available tasks PABLO_KIX_CMYB and SDOERR_BNB has been about double than GERARD_MO, all my 5 hosts received only GERARD_MO; not even one single task out of the other groups.

For me personally, it really doesn't make much of a difference which tasks are downloaded, as long as there are tasks available for download.

But in view of the above said, I would be interested, just out of curiosity, in knowing how the downloads are determined.
Anyone any idea?

Betting Slip
Send message
Joined: 5 Jan 09
Posts: 670
Credit: 2,498,095,550
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 46059 - Posted: 4 Jan 2017 | 12:00:28 UTC - in response to Message 46058.

They can and probably do give a priority setting to every job queue.

Bedrich has the right to abort the WU's he doesn't want to do but I think he may be stretching his rights a little by desiring to cherry pick the WU's that are offered him.

Maybe Bedrich should also note that the server regards an aborted WU the same way as an errored one thus reducing his MAX tasks per day. If he goes beyond 10 he will only be allowed 1 per day. Actually he will receive 2 per day because of the fact that computers count from 0, so 0=1 and 1=2

In my lowly opinion admins should let MAX tasks drop to 0 meaning you would get 1 until you had completed a valid WU.

PappaLitto
Send message
Joined: 21 Mar 16
Posts: 511
Credit: 4,617,042,755
RAC: 0
Level
Arg
Scientific publications
watwatwatwatwatwatwatwat
Message 46060 - Posted: 4 Jan 2017 | 12:41:12 UTC

The current server status shows over 11,000 long runs, over 2,000 are the GERARD_MO tasks, while over 9,000 are other tasks. Yet, the last few days, I have been getting mostly GERARD_MO tasks, which are slow and have low GPU usage, and a refuse to crunch them. When are the other tasks going to be released?

I have definitely noticed a job priority amongst the WUs. The Barnase Barstar WUs (BNB) definitely had the priority even though there were plenty of other WUs. I noticed the priority of these drop after 8000+ were finished, which makes sense, and yet there were still 2000-3000 unfinished BNB WUs. If you are heating your house, yes it is annoying to not have a consistent heat load but remember, you are legitimately contributing to real cutting edge biomedical research and I would not try to slow this down just because the utilization is slightly lower.

Erich56
Send message
Joined: 1 Jan 15
Posts: 1090
Credit: 6,603,906,926
RAC: 21,893,126
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 46061 - Posted: 4 Jan 2017 | 12:53:30 UTC - in response to Message 46060.

and yet there were still 2000-3000 unfinished BNB WUs.

The priority of the BNB WUs must have been lowered drastically - I have not received even one for almost a week now.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 46062 - Posted: 4 Jan 2017 | 15:45:12 UTC - in response to Message 46061.

and yet there were still 2000-3000 unfinished BNB WUs.

The priority of the BNB WUs must have been lowered drastically - I have not received even one for almost a week now.

I have 1 BNB running at the moment but the GERARDs seem to have the priority now (17 running). I've also noticed prioritization. I think the admins have mentioned it from time to time.

Lluis
Send message
Joined: 22 Feb 14
Posts: 26
Credit: 672,639,304
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 46068 - Posted: 4 Jan 2017 | 19:47:43 UTC - in response to Message 46062.

Job-----------------Type-------CPU sec.-----Cobblestones----Cobbl/second
Gerard_Nonaggre---Short---------16.747----------26.700----------1,59
Sdoerr_BNBs-------Long----------46.754---------492.000---------10,52
Gerard_Nonaggre---Short---------15.917----------26.700----------1,68
Gererd_MO_Mor2----Long----------33.908---------244.050----------7,20
Gerard_Nonaggre---Short---------16.790----------26.700----------1,59
Gerard_MO_TRV2----Long----------33.137---------244.050----------7,36
Gerard_MO_TRV2----Long----------34.350---------244.050----------7,10
Gerard_Nonaggre---Short---------16.827----------26.700----------1,59
Gerard_MO_TRV2----Long----------35.827---------244.050----------6,81
Gerard_MO_TRV2----Long----------46.682---------244.050----------5,23

My ten last jobs longer than 30 minutes.
As my goal is to help with cientific investigations I don't mind very much, but the ratio of Cobblestones/second is very variable.

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2343
Credit: 16,201,255,749
RAC: 7,520
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 46075 - Posted: 5 Jan 2017 | 11:46:29 UTC - in response to Message 46068.

My ten last jobs longer than 30 minutes:
Cobblestones Job Type CPU sec. Cobblestones /second Gerard_Nonaggre Short 16.747 26.700 1,59 Sdoerr_BNBs Long 46.754 492.000 10,52 Gerard_Nonaggre Short 15.917 26.700 1,68 Gererd_MO_Mor2 Long 33.908 244.050 7,20 Gerard_Nonaggre Short 16.790 26.700 1,59 Gerard_MO_TRV2 Long 33.137 244.050 7,36 Gerard_MO_TRV2 Long 34.350 244.050 7,10 Gerard_Nonaggre Short 16.827 26.700 1,59 Gerard_MO_TRV2 Long 35.827 244.050 6,81 Gerard_MO_TRV2 Long 46.682 244.050 5,23

As my goal is to help with scientific investigations I don't mind very much, but the ratio of Cobblestones/second is very variable.

The high difference in the ratio of Cobblestones/second between long and short workunits is intentional to motivate volunteers who has high-end GPUs to crunch long runs only, and let the volunteers who has middle range or low-end GPUs to crunch short runs (to meet the 1 day and 2 days quick return bonus, and the 5 days deadline). You can set it in your profile under "GPUGrid settings".
The difference in the ratio of Cobblestones/second within the same type of workunits is unintentional, as it comes from the inadequate method of estimating the flops needed to complete a workunit of a given batch.

Lluis
Send message
Joined: 22 Feb 14
Posts: 26
Credit: 672,639,304
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 46079 - Posted: 5 Jan 2017 | 18:18:40 UTC - in response to Message 46075.
Last modified: 5 Jan 2017 | 18:29:33 UTC


The high difference in the ratio of Cobblestones/second between long and short workunits is intentional to motivate volunteers who has high-end GPUs to crunch long runs only, and let the volunteers who has middle range or low-end GPUs to crunch short runs...


Thank you for your advice.
After 30 months crunching with a GTX 660 I didn't realize than uncheck the "Short" option in the preferences was a good service for the others crunchers now that I have a GTX 1070 (I thought that was selfish and after your information I realize is just the opposite).
By the way. Anybody can explain me what is Python Runtime ? Is aproppiate with my graphic card and Windows 10?

Tomas Brada
Send message
Joined: 3 Nov 15
Posts: 38
Credit: 6,768,093
RAC: 0
Level
Ser
Scientific publications
wat
Message 46086 - Posted: 6 Jan 2017 | 8:20:45 UTC - in response to Message 46079.


By the way. Anybody can explain me what is Python Runtime ? Is aproppiate with my graphic card and Windows 10?

It is describe in this news thread.
____________

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2343
Credit: 16,201,255,749
RAC: 7,520
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 46097 - Posted: 6 Jan 2017 | 23:56:28 UTC - in response to Message 46054.

The current server status shows over 11,000 long runs, over 2,000 are the GERARD_MO tasks, while over 9,000 are other tasks. Yet, the last few days, I have been getting mostly GERARD_MO tasks, which are slow and have low GPU usage, and a refuse to crunch them.
It took 2 days to crunch them. I don't think that it's so long that it should make anybody to selectively abort workunits, and propagate this bad attitude. I trust the staff that they know why they give higher priority to a given batch.

Why can we have the right to chose which task type, we want to crunch and which we don't want to crunch?
Besides it could have some positive effect on some crunchers, it would be a great chance for credit hunters, and it could have a negative impact on the project as a whole, as it would cause extra management work for the staff at the expense of scientific research.

When are the other tasks going to be released?
Judging by this question: if the task type selection would be realized, than you would flood the forum with requests for the workunits you prefer.

And lastly, if you are offended by my action and statement, I don't care!
What if I'm offended by that you don't care? :)

Bedrich Hajek
Send message
Joined: 28 Mar 09
Posts: 467
Credit: 8,190,146,966
RAC: 10,558,941
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 46101 - Posted: 7 Jan 2017 | 1:51:34 UTC - in response to Message 46097.

The current server status shows over 11,000 long runs, over 2,000 are the GERARD_MO tasks, while over 9,000 are other tasks. Yet, the last few days, I have been getting mostly GERARD_MO tasks, which are slow and have low GPU usage, and a refuse to crunch them.
It took 2 days to crunch them. I don't think that it's so long that it should make anybody to selectively abort workunits, and propagate this bad attitude. I trust the staff that they know why they give higher priority to a given batch.

Why can we have the right to chose which task type, we want to crunch and which we don't want to crunch?
Besides it could have some positive effect on some crunchers, it would be a great chance for credit hunters, and it could have a negative impact on the project as a whole, as it would cause extra management work for the staff at the expense of scientific research.

When are the other tasks going to be released?
Judging by this question: if the task type selection would be realized, than you would flood the forum with requests for the workunits you prefer.

And lastly, if you are offended by my action and statement, I don't care!
What if I'm offended by that you don't care? :)




I stand by my statements. I have a right to choose what I want to participate in. No one has the right to dictate to others.

Expressing an opinion that differs from the orthodoxy is not a bad thing. Following blindly is. It is good to keep the establishment on its toes.

And if you are offended by me and my opinions, then the problem is yours, not mine. People have to learn tolerance, especially of other people and things, they disagree with.

While my statements are provocative, they are by no means offensive, and I have no intention of involving myself in petty bickering. Though, I do enjoy lively debate. So, we disagree on a few things. Bad deal! Peace!!




Stefan
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 5 Mar 13
Posts: 348
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 46144 - Posted: 11 Jan 2017 | 14:51:04 UTC

Choosing which WUs to simulate opens a reaaaally nasty can of worms.
There were crunchers which wrote scripts to do that and we won't try to stop them but giving the option to all the users will cause a huge mess (plus would be hard for us to implement)

And yes, our WUs use priorities and BNBS is currently at top priority. Also apparently increasing the priority a lot tells BOINC to only send them to the best performing users but I didn't set it that high AFAIK.

Post to thread

Message boards : Server and website : Current Server Status

//