| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: Problems with the Due Time |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 49
|
|
| Author |
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
kinda same problem here, i installed boinc here at work to at least make my computer do something useful while programming. it's not even too slow - amd 64x2 3800 2gig of ram - but the work units are way too big. i only come around a few times a month and always for less than 6 hours. This effectively prevents me from completing any work units in time so atm running boinc at work is totally pointless and in fact more of a waste of power. Why the heck can't the work units be smaller?? I imagine a lot less power would be wasted by boinc users all over the world if wu unit would go down considerably, and i see no good reason not to do that. So please ne1 explain this to me it drives me nuts  ne1?, Yes, the larger units are mainly driven by load considerations. Don't want the whole server daemon complex to screech to a halt! It has. The size of the units is not a problem, it's the total number of transactions in a time segment. Rice jobs are fixed 8 hours regardless the speed of the computer and 12 days valid. Setting the client to go to sleep after say 3 hours idle post your 6 hours sessions and each time setting work fetch to no would allow to finish a unit, if you really want to. It can be tweaked to force the result to return before going cold to allow power savings to kick in if any. For the almost nuts to get a crunch, any. 
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at May 28, 2008 6:02:38 PM] |
||
|
|
mirabilos
Cruncher Joined: Mar 25, 2008 Post Count: 37 Status: Offline Project Badges: 
|
Speaking of rice jobs, they're all Valid (and one Pending Validation, but fine)
----------------------------------------on MirBSD. |
||
|
- |
zuriel
Cruncher Joined: Feb 20, 2007 Post Count: 3 Status: Offline |
ne1?, Yes, the larger units are mainly driven by load considerations. Don't want the whole server daemon complex to screech to a halt! It has. The size of the units is not a problem, it's the total number of transactions in a time segment. Well whatever cause the issue has, it would be could to be able to solve it. Rice jobs are fixed 8 hours regardless the speed of the computer and 12 days valid. ? Regardless of the speed? So, this amd64 x2 3800 machine here will get the workload it needs 8 hours for - and a 486 will download a tiny bit for which it also needs 8 hours? That's seems not well thought through. Is there a way to make my boinc do 3 hour jobs? this way my comp would not totally waste its processing power on never completing wus in time... Setting the client to go to sleep after say 3 hours idle post your 6 hours sessions and each time setting work fetch to no would allow to finish a unit, if you really want to. It can be tweaked to force the result to return before going cold to allow power savings to kick in if any. Well how do i force return what the comp has computed when i leave my work place? That'd work fine for me.. For the almost nuts to get a crunch, any. yeah nuts  and crunching |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
ne1?, Yes, the larger units are mainly driven by load considerations. Don't want the whole server daemon complex to screech to a halt! It has. The size of the units is not a problem, it's the total number of transactions in a time segment. Well whatever cause the issue has, it would be could to be able to solve it. Rice jobs are fixed 8 hours regardless the speed of the computer and 12 days valid. ? Regardless of the speed? So, this amd64 x2 3800 machine here will get the workload it needs 8 hours for - and a 486 will download a tiny bit for which it also needs 8 hours? That's seems not well thought through. Is there a way to make my boinc do 3 hour jobs? this way my comp would not totally waste its processing power on never completing wus in time... Setting the client to go to sleep after say 3 hours idle post your 6 hours sessions and each time setting work fetch to no would allow to finish a unit, if you really want to. It can be tweaked to force the result to return before going cold to allow power savings to kick in if any. Well how do i force return what the comp has computed when i leave my work place? That'd work fine for me.. For the almost nuts to get a crunch, any. yeah nuts and crunching Maybe if the techs see this they may have something to add to the 'not well thought through' observation. I dont'. A machine that runs 6 hours in a single session per week is better put on projects that have short jobs or at those that do allow to choose run time lengths. Latter AFAIK is only possible at Rosetta@Home. You could start your session at WCG by setting additional buffer / cache to 0.0000 and connect time to 0.001. That would pull a single unit per core. When that is done, switch the project to 'No New Tasks'. When done, it would send the result and go idle which would allow the machine to go into sleep mode. Presently HCC & DDDT are the shortest jobs averaging 7 hours. On my C2D they do about 6. ciao
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Jun 13, 2008 1:44:14 PM] |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
The biggest issue that BOINC has at the moment in regards to the concerns above is that it does not currently have the ability to distributed differently sized workunits to appropriate hosts. This means that powerful machines that are running BOINC 24x7 at 100% throttle with small queues will contact the server frequently to obtain new work and return completed work. Other computers such as a Pentium II or AMD K6 processor that only run BOINC for 30-45 minutes a day will find it almost impossible to complete any work at all.
The good news is that the BOINC clients for some time (at least 5.8 forward) return the following information about the client: % of time the computer is on % of time that BOINC is allowed to run research projects % cpu throttle estimated power of the computer (both the DCF and benchmark flops are available). This means that the information is available to obtain a reasonable guess as to how long (wall clock time) it will take for a computer to complete a given workunit. We have been working with BOINC to modify the BOINC scheduler so that it can take this data and then do some 'match making' between the workunits that are available to send, and the amount of power provided by a computer (using all of the metrics above). This will allow us to provide a range of sizes of workunits to distribute and the scheduler will be able to do a reasonable job assigning the work to the clients so that they can complete the work correct. This is a good way to deal with the problem becuase it means that we can send longer running workunits to the more powerful computers which are currently processing workunits in 3-4 hours while sending shorter workunits to weaker computers that are currently processing workunits in 24-36 hours (and due to only being on a portion of the day, they don't complete by the deadline). |
||
|
|
Diana G.
Master Cruncher Joined: Apr 6, 2005 Post Count: 3003 Status: Offline Project Badges: 
|
knreed, you are awesome
----------------------------------------  |
||
|
|
zuriel
Cruncher Joined: Feb 20, 2007 Post Count: 3 Status: Offline |
The biggest issue that BOINC has at the moment in regards to the concerns above is that it does not currently have the ability to distributed differently sized workunits to appropriate hosts. Exactly what i was trying to say We have been working with BOINC to modify the BOINC scheduler so that [...] the scheduler will be able to do a reasonable job assigning the work to the clients so that they can complete the work correct. Coool  That work is very much appreciated and i look forward to return a few WUs in time from work too.. |
|||
|
||||
|
mirabilos
Cruncher Joined: Mar 25, 2008 Post Count: 37 Status: Offline Project Badges:
|
Okay, time for some update:
----------------------------------------I’ve earned my rice badge almost (except 2 or 3 WUs) exclusively from running BOINC on MirBSD/i386 with the Linux emulation. This app does work superb. (And that where I personally prefer non-genfood…) The other apps are still not good; I’ll look at them next time they get an update and I get time to run BOINC and monitor it. |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
 Thanks for doing this, mirabilos. Lawrence |
||
|
|
|