Einstein running out of wu's on my machine

peristalsis
peristalsis
Joined: 20 Mar 05
Posts: 29
Credit: 49915332
RAC: 0
Topic 189808

Here is my problem. Normally run seti and einstein, but with seti being down, have been running einstein exclusively. Have einstein set for a six day connect time. But if I'm down to my last wu, and I'm online, einstein will not download additional work, even when using the update button. This means that if I'm away from this machine for 12 hours, and the wu finishes after one hour, the machine is idle for eleven hours. Am I missing something/doing something wrong? Thanks..p

Keith Moseley
Keith Moseley
Joined: 25 Aug 05
Posts: 6
Credit: 772
RAC: 0

Einstein running out of wu's on my machine

Quote:
Here is my problem. Normally run seti and einstein, but with seti being down, have been running einstein exclusively. Have einstein set for a six day connect time. But if I'm down to my last wu, and I'm online, einstein will not download additional work, even when using the update button. This means that if I'm away from this machine for 12 hours, and the wu finishes after one hour, the machine is idle for eleven hours. Am I missing something/doing something wrong? Thanks..p

I'm not sure but boinc may be down, I'm getting nothing from einstein or seti ( seti came up around 1am UK time this morning ) I got 3 w/u's from seti and they're done awaiting upload and 1 w/u awaiting d/l.

31/08/2005 16:26:28|Einstein@Home|Sending scheduler request to http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi
31/08/2005 16:26:28|Einstein@Home|Requesting 0 seconds of work, returning 0 results
31/08/2005 16:26:29|Einstein@Home|Scheduler request to http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi succeeded
31/08/2005 16:42:29|SETI@home|Started upload of 11my04aa.16929.23954.436076.163_3_0
31/08/2005 16:43:45|SETI@home|Temporarily failed upload of 11my04aa.16929.23954.436076.163_3_0: 500
31/08/2005 16:43:45|SETI@home|Backing off 3 hours, 54 minutes, and 3 seconds on upload of file 11my04aa.16929.23954.436076.163_3_0
31/08/2005 17:06:04|SETI@home|Started upload of 11my04aa.11069.27840.672170.194_2_0
31/08/2005 17:06:51|SETI@home|Temporarily failed upload of 11my04aa.11069.27840.672170.194_2_0: 500
31/08/2005 17:06:51|SETI@home|Backing off 3 hours, 10 minutes, and 49 seconds on upload of file 11my04aa.11069.27840.672170.194_2_0
31/08/2005 17:35:26|SETI@home|Started upload of 10my04aa.28802.22848.1034646.136_1_0
31/08/2005 17:36:42|SETI@home|Temporarily failed upload of 10my04aa.28802.22848.1034646.136_1_0: 500
31/08/2005 17:36:42|SETI@home|Backing off 59 minutes and 25 seconds on upload of file 10my04aa.28802.22848.1034646.136_1_0
31/08/2005 18:36:08|SETI@home|Started upload of 10my04aa.28802.22848.1034646.136_1_0
31/08/2005 18:36:27|SETI@home|Temporarily failed upload of 10my04aa.28802.22848.1034646.136_1_0: 500
31/08/2005 18:36:27|SETI@home|Backing off 3 hours, 19 minutes, and 9 seconds on upload of file 10my04aa.28802.22848.1034646.136_1_0

looks like boinc's down or having trouble to me, you can check my post " requesting 0 seconds of work? "

peristalsis
peristalsis
Joined: 20 Mar 05
Posts: 29
Credit: 49915332
RAC: 0

Thanks Keith. Maybe it is a

Thanks Keith. Maybe it is a feature in Boinc as I have a bunch of seti units waiting to upload if that is what is causing einstein to choke. Changed my prefs to 10 days and it still asks for 0 work. Maybe its just time to shut the machine off (blech). Always thought that was the idea of using Boinc..one project down...run another so no computer time is wasted. I read your other thread, and suspending/updating yielded the same zero request. thanks again..p

nfortino
nfortino
Joined: 7 Jun 05
Posts: 12
Credit: 1046710
RAC: 0

RE: Thanks Keith. Maybe it

Message 16088 in response to message 16087

Quote:
Thanks Keith. Maybe it is a feature in Boinc as I have a bunch of seti units waiting to upload if that is what is causing einstein to choke. Changed my prefs to 10 days and it still asks for 0 work. Maybe its just time to shut the machine off (blech). Always thought that was the idea of using Boinc..one project down...run another so no computer time is wasted. I read your other thread, and suspending/updating yielded the same zero request. thanks again..p

Actually, I think this a result of the scheduler in BOINC 4.45. It attempts to insure the user's CPU share is honored in the long run, while also meeting deadlines. For the long run, it employs a "long term debt," which is basically the amount of CPU time a project should have gotten minus the amount it actually got. Projects with greater LTD download first, and projects with negative LTD are not allowed to DL work. While this works well on paper, in practice there are complications. Since SETI was down for about a week, einstein took more than its share of CPU, causing SETI to build up LTD, and einstein to have a significantly negative LTD. Thus, when SETI came back online, it had priority, and einstein was locked out (and will be for a while). Personally, I believe if a project is down for a long time, it should loose its CPU time, but I am not sure how that could be implemented.

peristalsis
peristalsis
Joined: 20 Mar 05
Posts: 29
Credit: 49915332
RAC: 0

It could be. I fiddled

It could be. I fiddled around with percentages (90% einstein/10%seti) and boinc gave me the following nice speech:
"09/01/05 12:59:34|SETI@home|Scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded
09/01/05 12:59:34|SETI@home|Message from server: No work sent
09/01/05 12:59:34|SETI@home|Message from server: (won't finish in time) Computer on 96.0% of time, BOINC on 99.9% of that, this project gets 9.1% of that
09/01/05 12:59:34|SETI@home|No work from project
09/01/05 12:59:35|SETI@home|Deferring communication with project for 1 days, 23 hours, 59 minutes, and 58 seconds"
Of course its still trying to d/l another dozen units with 'no work' and 'project suspended' selected. And no uploading of my 15 reday wu's. I intended to idle seti until everything got caught up but I guess boinc doesn't believe in that. Hopefully the next revision will have this time prioritization sub-routine working properly. If things aren't OK by tomorrow afternoon I'm going to edit the appropriate file(s) and zero all debts.

Winkle
Winkle
Joined: 24 Aug 05
Posts: 8
Credit: 47604
RAC: 0

RE: Actually, I think this

Message 16090 in response to message 16088

Quote:
Actually, I think this a result of the scheduler in BOINC 4.45. It attempts to insure the user's CPU share is honored in the long run, while also meeting deadlines. For the long run, it employs a "long term debt," which is basically the amount of CPU time a project should have gotten minus the amount it actually got. Projects with greater LTD download first, and projects with negative LTD are not allowed to DL work. While this works well on paper, in practice there are complications. Since SETI was down for about a week, einstein took more than its share of CPU, causing SETI to build up LTD, and einstein to have a significantly negative LTD. Thus, when SETI came back online, it had priority, and einstein was locked out (and will be for a while). Personally, I believe if a project is down for a long time, it should loose its CPU time, but I am not sure how that could be implemented.

This LTD would explain why Seti has D/L a heap of work on my machine while Einstein finished what it was doing and now refuses to D/L any more. I wonder if there is a way of cancelling the LTD count ?

Edit:
I just looked at the client state files, and these seem to explain it.

C:\Program Files\BOINC\client_state.xml

Machine 1 (Both Run, as the debts are the same)
project_name>Einstein@Home
short_term_debt 1047.926895
long_term_debt -15404.039895

project_name SETI@Home
short_term_debt 1047.926895
long_term_debt -15404.039895

Machine 2 (Einstein won't run as the debts are different, one is positive)
project_name Einstein@Home
short_term_debt 0.000000
long_term_debt -28618.382787

project_name SETI@Home
short_term_debt 0.000000
long_term_debt 28618.382787

RandyC
RandyC
Joined: 18 Jan 05
Posts: 6616
Credit: 111139797
RAC: 0

RE: This LTD would explain

Message 16091 in response to message 16090

Quote:
This LTD would explain why Seti has D/L a heap of work on my machine while Einstein finished what it was doing and now refuses to D/L any more. I wonder if there is a way of cancelling the LTD count ?

One way to clear the LTD is to reset the project. Do this only if:
1) All WU have completed and uploaded
2) All WU have been reported to the Project
3) No WU are queued for D/L

If the above is not true, you will lose whatever is pending when you do the reset. You may have to reset the project several times before the LTD is down near zero.

Seti Classic Final Total: 11446 WU.

Metod, S56RKO
Metod, S56RKO
Joined: 11 Feb 05
Posts: 135
Credit: 826639986
RAC: 84338

RE: Personally, I believe

Message 16092 in response to message 16088

Quote:
Personally, I believe if a project is down for a long time, it should loose its CPU time, but I am not sure how that could be implemented.

My feelings exactly. Perhaps it is sensible to let a project accumulate some LTD but not too much. Perhaps a couple of CPU days. SO perhaps there should be a ceiling (and floor?) of LTD, preferably user-configurable.

Metod ...

5 and a half of 13
5 and a half of 13
Joined: 23 Feb 05
Posts: 36
Credit: 21194
RAC: 0

Can't you just set Seti not

Can't you just set Seti not to download WUs? Then you will download Einstein WUs no problem :)

Won't you be able to pay off the long term debt as slowly or fast as you want to then? Or will the long term debt buildup actually get worse if you keep on switching other projects on and off?

PS: If I have the wrong end of the stick please just roast me lightly :)

Need Help? Try the excellent Unofficial BOINC Wiki!
We are the BOINC. Prepare to be assimilated.
'anthrax beats WinXP' - The Register

peristalsis
peristalsis
Joined: 20 Mar 05
Posts: 29
Credit: 49915332
RAC: 0

Originally I had seti

Originally I had seti suspended & no new work, but it still tried to upload and download. Problem was if I had only one einstein unit in my queue, boinc would not d/l any additional einstein units. With a dial-up connection, this meant the machine would be idle for the time between finishing the einstein unit and my next access to the 'net. If I had an einstein unit needing an hours runtime to complete, and I would be away from the computer for a couple of days, the unit would complete and then dead time. Connect times (one day/ten days) had no affect on this. Conjecture was that the backlog of seti wu's was causing boinc to think I had alot of work needing to run in my other project, counted that as part of the long term debt and just wouldn't get more einstein units. Seti is now working well enough that I have work so I'm no longer worried. But in actually why should I worry about a trivial (gasp) computer program when there is a sufficiency of important things in the world to concern me...p

5 and a half of 13
5 and a half of 13
Joined: 23 Feb 05
Posts: 36
Credit: 21194
RAC: 0

RE: Originally I had seti

Message 16095 in response to message 16094

Quote:
Originally I had seti suspended & no new work, but it still tried to upload and download. ......

Sorry Peristalsis, I didn't read your message 17759 properly the first time or I wouldn't have raised a redundant point.

Need Help? Try the excellent Unofficial BOINC Wiki!
We are the BOINC. Prepare to be assimilated.
'anthrax beats WinXP' - The Register

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.