Computers that hoard work units and return none

voidxor
voidxor
Joined: 16 Sep 05
Posts: 12
Credit: 22168953
RAC: 0
Topic 192115

I've been chugging quietly here for awhile, but today I noticed that a couple results that my computers completed over a week ago are still pending credit because the following two computers are hoarding work units:

  • *

http://einsteinathome.org/host/760296
*http://einsteinathome.org/host/796471

Why are these machines taking so many work units without returning any? Why is the work scheduler even giving these machines that much work? Is this normal or is something wrong?

Nothing But Idle Time
Nothing But Idl...
Joined: 24 Aug 05
Posts: 158
Credit: 289204
RAC: 0

Computers that hoard work units and return none

I first computer 760296 had only short tasks when I randomly checked; the avg turnaround time for it is 8 days. That combination would require a large number of tasks on hand.

The second individual had only 5 tasks outstanding when I checked and that is no problem.

Dimmerjas
Dimmerjas
Joined: 6 Jul 05
Posts: 28
Credit: 343971
RAC: 0

We all have 14 days to return

We all have 14 days to return results for WU's before they reach deadline. Some users might run other project's along with E@H so they need close to 14 days to return results, or my not have there computer(s) running BOINC-projects each and every day. The first computer, 760296, returned results the day before yesterday. And the next deadline is in 3 days. So results for those WU's, might be returned in time?
As for the second computer, 796471, those results may not be returned in time? But then the WU's will be send to another user, that might finish them in no time.

I have, many times, had more than 1000 Pending credits in my account, but the next day, Pending credits may have dropped to 600-700, becouse my workmates have returned results.

So keep on going. And don't worry. You will get the credits for your contribution to this project.

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494410
RAC: 0

Multiple projects on a

Multiple projects on a Coppermine? Doesn't sound like a good idea to me. Of course, it's possible, but I'd say that kind of box has more than enough to do crunching one project... and it looks like the turnaround time proves me right there.

Dimmerjas
Dimmerjas
Joined: 6 Jul 05
Posts: 28
Credit: 343971
RAC: 0

RE: Multiple projects on a

Message 54508 in response to message 54507

Quote:
Multiple projects on a Coppermine? Doesn't sound like a good idea to me. Of course, it's possible, but I'd say that kind of box has more than enough to do crunching one project... and it looks like the turnaround time proves me right there.

You'r right! - Maby not a good idea, but multiple projects on such a box is possible. It may not run multiple project's "side by side". One projec't can be set on "Suspend", while another project is running, and visa versa. What do I know?
This box does not belong to a "nubie". The user have 6 box's asigned to E@H, an account with 927,549.84 Credits, and RAC 1,830.51. All 6 box's have contacted the project-server within the last 2 days. Some of them today. So I would say this is a dedicated user, that most likely will return results within deadline.

tekwyzrd
tekwyzrd
Joined: 25 Feb 05
Posts: 49
Credit: 2922090
RAC: 0

RE: Multiple projects on a

Message 54509 in response to message 54507

Quote:
Multiple projects on a Coppermine? Doesn't sound like a good idea to me. Of course, it's possible, but I'd say that kind of box has more than enough to do crunching one project... and it looks like the turnaround time proves me right there.

I guess it depends on the computer and the individual. I run a dual 700 MHz coppermine computer that does work for eight projects. The only results never reported were due to my actions, an example being a failed upgrade to the newest SuSE version or problems with hardware like a bad bios flash. No problem completing and returning results within the allotted time.

Nothing travels faster than the speed of light with the possible exception of bad news, which obeys its own special laws.
Douglas Adams (1952 - 2001)

voidxor
voidxor
Joined: 16 Sep 05
Posts: 12
Credit: 22168953
RAC: 0

Thank you for all your quick

Thank you for all your quick responces. I have a broadband connection and have my "Connect to network about every 'x' days" setting left at .1 days, so I am not used to seeing more than a couple work units handed out to a given computer at a time.

I realise now that the "Connect to network about every 'x' days" setting and "Average turnaround time" value are the issues here. Upon reading the BOINC Wiki's article on the work buffer and the explaination of average turnaround time, I understand that this is normal behavior for a properly configured computer on a dial-up connection.

Please note that I'm not really here for credit, I'm here for the science. I just started this thread because I thought those two computers were having problems. I came to BOINC from Grid.org a little over a year ago, and if you are familiar with that project (and it's problems) at all, you'll understand why I'm so paranoid about work units falling into black holes. :-P

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494410
RAC: 0

Well, personally, I prefer a

Well, personally, I prefer a small cache aswell. You have a better chance of getting credit quickly and so do your fellow crunchers. Besides, it just looks neater in the manager ;-) makes it easier to get an overview. And (main reason) you never get deadline issues when a project completely underestimates the time a WU needs to finish (had that with HashClash a few times; WUs have very different sizes there... mine were supposed to take 2 hours apiece, well, they took 8 or 10... I let my PC run overnight and still couldn't get everything done, which cured me from storing a day or more of WUs in my cache if I can avoid it).
But I've never tried participating in BOINC on dial-up. For that I guess it's really better to have a larger cache. Besides, there are much more extreme examples out there then the hosts you picked ;-)

Nothing But Idle Time
Nothing But Idl...
Joined: 24 Aug 05
Posts: 158
Credit: 289204
RAC: 0

RE: Well, personally, I

Message 54512 in response to message 54511

Quote:
Well, personally, I prefer a small cache as well. You have a better chance of getting credit quickly and so do your fellow crunchers. (...)

Very small cache is good when everything works well, yet each must tailor to his/her own environment. With EaH going offline recently and the fact that I've experienced 3 ISP outages ranging from 1/2 to 1.x days I've capitulated and increased my que from a long standing 0.01 to 1 day, then backed off slightly to 0.7. This que size seems to work for me and I can produce a satisfactory turnaround.

Erik
Erik
Joined: 14 Feb 06
Posts: 2815
Credit: 2645600
RAC: 0

I keep mine in the 1 to 2 day

I keep mine in the 1 to 2 day range. I usually do good to get the results turned in in a timely manner with the exception of one borrowed computer that occasionally gets the communication turned off or BOINC shut down. So I try to keep an eye on it through the computer tabs in the BOINC manager and get the results turned in before time runs out.

Huff
Huff
Joined: 5 Jan 06
Posts: 36
Credit: 1378476
RAC: 0

I have found that 1 day works

I have found that 1 day works well, however the big machine (668232)ran out of work (the half hour 12 point units). There were hundreds of them waiting to be reported before the file server came back on line. That had me a little concerned. It sat Idle for a few hours now it is back up to 3600+ pending credits, it usually runs about 2500 to 4000 pending.

Comment viewing options

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