Priority setting |
Message boards : Preferences : Priority setting
Author | Message |
---|---|
Is there a parameter control to set run priority please? |
|
ID: 6345 | Rating: 0 | rate: / | |
Is there a parameter control to set run priority please? You need to reduce your \'cache\' size to something less than about 1.75 days. This is due to the way BOINC calculates which wu\'s it needs to work on to ensure the work is completed before the work unit deadline is reached. And since Malaria Control work units have always had a 3 day deadline this part isn\'t going to change. Live long and BOINC! ____________ Paul (S@H1 8888) |
|
ID: 6347 | Rating: 0 | rate: / | |
Thanks, will do that. |
|
ID: 6349 | Rating: 0 | rate: / | |
I\'m having the same problems. I\'ve got my cache set to to 3 days and I also always get malariacontrol.net WU\'s in priority as soon as some of them are downloaded. |
|
ID: 6446 | Rating: 0 | rate: / | |
I don\'t want to be stubborn, but why would you not increase your deadline to a week or so? Rather than rewrite or paraphrase what Nick wrote, I\'ll just link to it here: work creation and deadlines. |
|
ID: 6449 | Rating: 0 | rate: / | |
Rather than rewrite or paraphrase what Nick wrote, I\'ll just link to it here: work creation and deadlines. Thanks for pointing me to the explanation. At least now I know why it is they use these short deadlines, though I\'m not yet convinced it\'s really necessary. The explanation says the imbalances will settle out in the long term, but with me it results in many WU\'s (also of another project I\'m running in that computer, proteins, which also has not too long deadlines) not reaching their deadlines. I like to use more than one day of cache because I\'ve had hardware failure (router/cable modem) problems in the past and it can take a couple of days to replace them, and I don\'t want to waste too many potential flops in the mean time if it would happen again. On another computer I\'ve got continuously 7 projects running without ever running into deadlines and without any of them giving me problems with my cache-size. I still don\'t understand why malariacontrol has to be so different. If they want as much results as possible, they shouldn\'t bug their crunchers with WU-management they shouldn\'t be bothered with (that\'s what the Boinc platform is supposed to do). If they don\'t care about people with larger cache sizes than 1.75 days running into deadlines and consequently deleting WU\'s that pass the deadlines, than things are fine as they are of course. It\'s not what I like to do, but I don\'t want to decrease my cache size either, more like the other way around, still because of Malariacontol I don\'t. Well, hopefully some day they\'ll find a solution. Thanks for your reply anyway. I/I |
|
ID: 6451 | Rating: 0 | rate: / | |
Thanks for pointing me to the explanation. At least now I know why it is they use these short deadlines, though I\'m not yet convinced it\'s really necessary. :) That\'s where the project admins have to come in and convince you. The explanation says the imbalances will settle out in the long term, but with me it results in many WU\'s (also of another project I\'m running in that computer, proteins, which also has not too long deadlines) not reaching their deadlines. If they want as much results as possible, they shouldn\'t bug their crunchers with WU-management they shouldn\'t be bothered with (that\'s what the Boinc platform is supposed to do). The scheduling and debt experts need to address these concerns, however I can say that while BOINC is still not perfect in wu management (will it ever be?), it\'s improved vastly over the years. I also suspect that in your case it may need to learn a bit more about your situation over the long term and that means weeks or even a couple of months, not days. These days, micromanagement tends to interfere with, rather than aid boinc\'s learning ability. This may mean that you need to let it miss a few deadlines and make the necessary debt adjustments before it holds back on the work requests. I understand this may seem counter intuitive and may not sit well, however it may be what is required. I suspect that cancelling the wu\'s before they\'ve been crunched prevents boinc from learning that it\'s overcommitted and allowing it to make the necessary adjustments. I have used a few suspicions here because I use a small cache and concentrate my work mainly on this project, so fortunately (unfortunately?) I don\'t experience these problems. If they don\'t care about people with larger cache sizes than 1.75 days running into deadlines and consequently deleting WU\'s that pass the deadlines, than things are fine as they are of course. I find that the admins here do actually care for their contributors. Far more than at other projects, which is why we have such a casual, relaxed and positive relationship with them. I don\'t believe that retaining their existing policies regarding deadlines is a question of not caring. You\'ve read the official reasoning that, as far as I can tell, still holds. As for deleting wu\'s, the system is designed with redundancy to cater for such situations, so the work that you discard will be recycled without much fuss. TBH, I suspect you will eventually tire of this micromanagement behaviour because it seems like effort best spent on more productive pursuits. :) |
|
ID: 6453 | Rating: 0 | rate: / | |
If it may take month to settle, the complete debt-system would need re-pavement, but I think this isn\'t the case. Though I\'m only crunching for a couple of month, my crunching habits are pretty consistent, so there would be no need for the learning system to get out of control all the time. On my other computer this is also not necessary, but I don\'t run malariacontrol on that one so it can\'t disturb the scheduler. But as far as my understanding of the debt system goes, it doesn\'t need to take that long, and deleting WU\'s to keep it out of priority mode only speeds up the settlement process, but it has to be repeated once in a while, and when attaching new projects, or running projects that do not deliver WU\'s all the time, like LHC. |
|
ID: 6455 | Rating: 0 | rate: / | |
Boinc keeps on crunching past-deadline WU\'s (I don\'t understand why it keeps on crunching over-date WU\'s in the first place) and other projects also will pass their deadlines that otherwise wouldn\'t have done that, so I\'ll have to micromanage that anyway not to lose too many flops (and I assume projects don\'t need the over-date WU\'s anymore as well). I/I In SOME instances, a unit that is returned after its deadline will still get credit. These are pretty specific but they are: the unit is re-issued to someone else because it did not get retuned on time, that second computer does not return that unit before the original computer, that is still crunching the same unit, the original computer will get credit issued to it. If the second computer does return the unit before the original computer, then the original computer will get no credits. When I was a message board admin over at Seti we saw this question many times and they, being the writers of Boinc, were working on setting up the scheduler part so that the overdue units were sent out to subsequent computers that consistently returned units in a very short period of time. That means that if you have not started a unit before it has expired and then return it almost immediately after crunching that unit, you probably will not get credit for that unit. I do however agree with Chris in that your process of aborting the unit is not letting Boinc figure out that it is getting too much work for your way of crunching and you may be micro-managing for a very long time to come. Each Project has its own settings for how much work to get and maintain, Your Account, Computing Preferences, Computer is connected to the Internet about every (Leave blank or 0 if always connected. BOINC will try to maintain at least this much work.). If you set this to say 1 day, then you will have 3 days to crunch that one days worth of work and maybe Boinc can figure it out for you, without causing you to crunch for nothing. ____________ |
|
ID: 6472 | Rating: 0 | rate: / | |
Mikey, that\'s the way it\'s supposed to happen and it does happen that way most of the time but too often it just doesn\'t happen that way. The reasons why it doesn\'t always happen that way are well known so I\'m not going to rehash that tale here. |
|
ID: 6489 | Rating: 0 | rate: / | |
I\'ve been having too many problems all the time with this deadline issue and unfortunately I had to stop crunching this project. If the project can\'t find a solution for this, then I won\'t for sure. I never crunch for the credits (I don\'t understand those who do), but I just get too many problems with other projects running on the same machine that are getting pushed over their deadline. And all of this because I\'ve set a 3-day cache. |
|
ID: 6808 | Rating: 0 | rate: / | |
I\'ve been having too many problems all the time with this deadline issue and unfortunately I had to stop crunching this project. If the project can\'t find a solution for this, then I won\'t for sure. I never crunch for the credits (I don\'t understand those who do), but I just get too many problems with other projects running on the same machine that are getting pushed over their deadline. And all of this because I\'ve set a 3-day cache. I think the \'problem\' is that the data here at Malaria is used by the scientists fairly quickly after it is returned by us crunchers. Other Projects, Seti for example, has data that was crunched years ago and is still not being analyzed. That is just one of the reasons I left and will never go back! Anyway with the scientists here inputting our results into their work fairly quickly, longer turnaround times are probably not in the plans. Some projects just do not play nice with each other on some computers schedules. Apparently yours is one of them. If in the future things change, on either side, you will be welcomed back with open arms!!! ____________ |
|
ID: 6812 | Rating: 0 | rate: / | |
Mickey, you may hug me as I\'ve embraced Malariacontrol again. I\'ve been shifting around with some of the projects between my computers. My other machine only had pretty long WU\'s from other projects and because of that only one or two Malariacontrol WU\'s are download at a time after I shifted Malariacontrol to that one. I\'ve been testing for about a week and so far I haven\'t run into deadline problems anymore. |
|
ID: 6935 | Rating: 0 | rate: / | |
Mickey, you may hug me as I\'ve embraced Malariacontrol again. I\'ve been shifting around with some of the projects between my computers. My other machine only had pretty long WU\'s from other projects and because of that only one or two Malariacontrol WU\'s are download at a time after I shifted Malariacontrol to that one. I\'ve been testing for about a week and so far I haven\'t run into deadline problems anymore. Welcome back! Good to have you back. The more the merrier. |
|
ID: 6937 | Rating: 0 | rate: / | |
Mickey, you may hug me as I\'ve embraced Malariacontrol again. I\'ve been shifting around with some of the projects between my computers. My other machine only had pretty long WU\'s from other projects and because of that only one or two Malariacontrol WU\'s are download at a time after I shifted Malariacontrol to that one. I\'ve been testing for about a week and so far I haven\'t run into deadline problems anymore. A virtual hug is headed your way! ____________ |
|
ID: 6967 | Rating: 0 | rate: / | |
Blush :) |
|
ID: 6969 | Rating: 0 | rate: / | |
Message boards : Preferences : Priority setting