Pending units


log in

Advanced search

Message boards : Number crunching : Pending units

1 · 2 · 3 · 4 . . . 8 · Next
Author Message
Miklos M.
Send message
Joined: 5 May 99
Posts: 755
Credit: 15,601,703
RAC: 14,630
United States
Message 995775 - Posted: 12 May 2010, 13:12:36 UTC

Looks like my pending units keep on piling up without credit. I wonder if something "broken" at SETI.
____________

Profile James SotherdenProject donor
Avatar
Send message
Joined: 16 May 99
Posts: 8814
Credit: 34,489,497
RAC: 58,672
United States
Message 995826 - Posted: 12 May 2010, 16:30:18 UTC - in response to Message 995775.


What is the amount you normally have?
As of today I have 60,000 pendings not unusual. I hover around 50 to 60 thousand on average.

____________

Old James

WinterKnight
Volunteer tester
Send message
Joined: 18 May 99
Posts: 8644
Credit: 24,176,257
RAC: 20,940
United Kingdom
Message 995833 - Posted: 12 May 2010, 16:55:52 UTC

I think the agreed position is that pendings in most accounts is at least three times RAC.

Mine at the moment is about 4 * RAC. But it can vary quite a bit depending on server status, size of cache etc.

Sten-Arne
Volunteer tester
Send message
Joined: 1 Nov 08
Posts: 3406
Credit: 20,091,522
RAC: 22,196
Sweden
Message 995844 - Posted: 12 May 2010, 17:58:31 UTC - in response to Message 995833.

I think the agreed position is that pendings in most accounts is at least three times RAC.

Mine at the moment is about 4 * RAC. But it can vary quite a bit depending on server status, size of cache etc.


Mine is at the moment 4,77744 * RAC

woodenboatguy
Send message
Joined: 10 Nov 00
Posts: 368
Credit: 3,969,364
RAC: 0
Canada
Message 995876 - Posted: 12 May 2010, 21:29:59 UTC

6.4 and change. I had one of the crunchers start failing on a graphics card so it's been offline occasionally and that's dropped the RAC.

On a previous thread we all concluded something between 4 and 6 was pretty standard.

Regards,
____________

Miklos M.
Send message
Joined: 5 May 99
Posts: 755
Credit: 15,601,703
RAC: 14,630
United States
Message 995900 - Posted: 12 May 2010, 23:39:02 UTC - in response to Message 995826.

My ratio used to be about 1-2 times RAC, now it is 4 times and climbing.
____________

Profile Michael Belanger
Avatar
Send message
Joined: 30 Jul 00
Posts: 1615
Credit: 1,764,570
RAC: 2,187
United States
Message 995922 - Posted: 13 May 2010, 3:27:34 UTC
Last modified: 13 May 2010, 3:29:11 UTC

I'm currently at 3.966799507 (or thereabouts). For whatever reason, my RAC seems to be fluctuating somewhat lately; same (kind of) problem as Miklos; my Pendings seem to be piling up - maybe the others doing the same units are just taking longer?
____________

Profile Francesco Forti
Avatar
Send message
Joined: 24 May 00
Posts: 281
Credit: 138,787,067
RAC: 49,731
Switzerland
Message 995943 - Posted: 13 May 2010, 9:22:34 UTC

I'm currently at 5.1350
My quicker Host (8 CPU + 260 GPU) now have a pending / RAC ratio of 5.3323
I started it on 22 April and now has 80'000 Pending, 15'000 RAC and 343'628 done.
This host has a Average turnaround time of 5.35 days. After 21 days this host should be near at the normal productivity RAC (I have calculated about 20~22K RAC) but it is still at 15'000 because - I suppose - of the 80'000 pending.

In the server status page I see also that "Results out in the field" is 6,185,869
This is very high! I think it's about double than some week or month ago but I don't have any way to be sure.
See also Result turnaround time (last hour average) 122.13 hours (I remember it was less but I'm not sure). It is about 5 days so my host's turnaround time is syncronized with the global average time.

I think that or a lot of people have increased the work buffer in last months or something is happening in the project.

At last, I see the HOSTs with GPU have higher Pending/RAC Ratio.

Bye,
Franz



____________

Profile Francesco Forti
Avatar
Send message
Joined: 24 May 00
Posts: 281
Credit: 138,787,067
RAC: 49,731
Switzerland
Message 995979 - Posted: 13 May 2010, 15:03:29 UTC

Yust one idea:
Most of GPU users are running the optimized "V12 modification by Raistmer".
So the "VLAR autokill enabled" is ON.
Some of them (as I) are using also the Reschedule 1.9 program.
In this case I don't "autokill" anything and I crunch all WU's I receive.
I move from CPU to GPU some work ad move VLAR from GPU to CPU.
But if other crunchers autokills the same VLAR of the same Workunit, I'll wait a long time (in progress until deadline) untill someone doesn't autokill the VLAR that I have crunched at next reassignement.
This can explain high pending and growing "Results out in the field".

What do you think?

Franz


____________

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 15,571,963
RAC: 11,458
United States
Message 995981 - Posted: 13 May 2010, 15:24:40 UTC - in response to Message 995979.

Another problem is people quitting and leaving WUs to sit until they time out. Whenever SETI has a problem we seem to lose a lot of people. A lot of my pending is from right around the time when there were problems.

On another note, it might be an idea to make it easier for people to quit. I had one machine I shut down that left behind a few WUs. I did detach first but did not reattach so the servers never got the word that machine was gone. If there was a button that could be used to say hey, I'm leaving, thanks for all the fish....uhh credits, then we might be able to get whatever WUs they have left on their machines back out in the field and crunched a lot sooner.
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile James SotherdenProject donor
Avatar
Send message
Joined: 16 May 99
Posts: 8814
Credit: 34,489,497
RAC: 58,672
United States
Message 995988 - Posted: 13 May 2010, 16:22:04 UTC

Id have to say That I agree with Francesco that a lot of cruncehers upped their cache to ten days. Me I run a 0.75 day cache. Even my old P4 has 20 pendings.

Yes I run out of work once in awhile, But thats when you blow out the dust bunnies.
____________

Old James

Profile Fred J. Verster
Volunteer tester
Avatar
Send message
Joined: 21 Apr 04
Posts: 3242
Credit: 31,787,810
RAC: 4,541
Netherlands
Message 995990 - Posted: 13 May 2010, 16:33:00 UTC - in response to Message 995988.
Last modified: 13 May 2010, 16:46:45 UTC

Mine is 5.6809105 (x RAC) and a 3 day's cache, more then enough, especially with a few projects all together.
Just started MW on ATI-GPU's.
____________

Profile Gundolf Jahn
Send message
Joined: 19 Sep 00
Posts: 3184
Credit: 358,489
RAC: 28
Germany
Message 996012 - Posted: 13 May 2010, 17:50:22 UTC - in response to Message 995979.

But if other crunchers autokills the same VLAR of the same Workunit, I'll wait a long time (in progress until deadline) untill someone doesn't autokill the VLAR that I have crunched at next reassignement.

Why would you have to wait for a deadline? If a task is killed, it's reported back immediately and gets sent out again. The only thing you'll have to wait for is that it works its way up someone's (possibly ten-day) cache.

Gruß,
Gundolf
____________
Computer sind nicht alles im Leben. (Kleiner Scherz)

SETI@home classic workunits 3,758
SETI@home classic CPU time 66,520 hours

Jim WelchProject donor
Send message
Joined: 17 May 99
Posts: 16
Credit: 6,993,125
RAC: 1,220
United States
Message 996064 - Posted: 14 May 2010, 0:23:57 UTC - in response to Message 996012.

This discussion raises a question I have wondered about. What is the reason for GPU users to need/want an autokill feature for VLAR workunits? Unless these workunits don't function correctly with GPU crunchers for some reason, it seems to me a credit is a credit. Does it really make a difference between running a lot of low credit workunits as opposed fewer high credit workunits? And yes, I am a CPU cruncher only.
____________

Profile Paul D. Buck
Volunteer tester
Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 996069 - Posted: 14 May 2010, 1:17:24 UTC - in response to Message 996064.

This discussion raises a question I have wondered about. What is the reason for GPU users to need/want an autokill feature for VLAR workunits? Unless these workunits don't function correctly with GPU crunchers for some reason, it seems to me a credit is a credit. Does it really make a difference between running a lot of low credit workunits as opposed fewer high credit workunits? And yes, I am a CPU cruncher only.

Yes, credit is credit... but, in essence each SaH task pays a semi-fixed rate, and the VLAR tasks take 3-10 times longer than the other tasks, so, you get 1/3 to 1/10 the going rate per second by doing the VLAR on the GPU ... it is like using a sports car to haul a pile of dirt ... you can get there, but it would be faster with a truck ...

So, by killing the long running tasks that take an hour to an hour and a half to run on a GPU and concentrating on those tasks that only take 5-10 minutes to run on the same GPU, work per hour (RAC is a crude measure of this) is much higher... we call that efficiency ...

Profile Francesco Forti
Avatar
Send message
Joined: 24 May 00
Posts: 281
Credit: 138,787,067
RAC: 49,731
Switzerland
Message 996105 - Posted: 14 May 2010, 6:53:29 UTC - in response to Message 996012.


Why would you have to wait for a deadline? If a task is killed, it's reported back immediately and gets sent out again. The only thing you'll have to wait for is that it works its way up someone's (possibly ten-day) cache.

Gruß,
Gundolf

Sure? You are right, and the system should act so but I'm not sure at all.
I think that resend of killed task to other cruncher is done after the weekly maintenance.
Can some one tell us haw the real procedure is?

By the way: I see
Results returned and awaiting validation 6,043,837
Results out in the field 6,284,723

Why are awaiting for validation?
____________

Josef W. SegurProject donor
Volunteer developer
Volunteer tester
Send message
Joined: 30 Oct 99
Posts: 4247
Credit: 1,048,692
RAC: 265
United States
Message 996147 - Posted: 14 May 2010, 11:45:10 UTC - in response to Message 996105.


Why would you have to wait for a deadline? If a task is killed, it's reported back immediately and gets sent out again. The only thing you'll have to wait for is that it works its way up someone's (possibly ten-day) cache.

Gruß,
Gundolf

Sure? You are right, and the system should act so but I'm not sure at all.
I think that resend of killed task to other cruncher is done after the weekly maintenance.
Can some one tell us haw the real procedure is?

When a reissue is created it goes at the end of the "Results ready to send" queue, so for this project under normal conditions is sent out within a few hours.

By the way: I see
Results returned and awaiting validation 6,043,837
Results out in the field 6,284,723

Why are awaiting for validation?

Mostly because the wingmate's result has not yet been received. That's the total "pending", some may be inconclusives too.
Joe

Profile Francesco Forti
Avatar
Send message
Joined: 24 May 00
Posts: 281
Credit: 138,787,067
RAC: 49,731
Switzerland
Message 996190 - Posted: 14 May 2010, 15:52:04 UTC - in response to Message 996147.


When a reissue is created it goes at the end of the "Results ready to send" queue, so for this project under normal conditions is sent out within a few hours.


Ok, I have seen now that you are right.
After 1 hour the task was reassigned to a new user.
http://setiathome.berkeley.edu/workunit.php?wuid=610657396


Mostly because the wingmate's result has not yet been received. That's the total "pending", some may be inconclusives too.

Ok, clear! But why it's so high? Weeks ago it wasn't so.
Francesco

____________

Profile Paul D. Buck
Volunteer tester
Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 996197 - Posted: 14 May 2010, 16:31:01 UTC - in response to Message 996190.

...But why it's so high? Weeks ago it wasn't so.

probably because I started doing SaH work again ... I broke every last little thing ...

1 · 2 · 3 · 4 . . . 8 · Next

Message boards : Number crunching : Pending units

Copyright © 2014 University of California