Now I am running a number of projects and boinc seems to think it is clever to keep switching between them. Well actually NO ! that is not a clever idea and I don't want it to but can't stop it save for manually micromanaging the software. I want a task to start and run until complete, not switch to another taks and leave one partly done. I can't always guarantee how long the Pc will be on for and I really don't see the point in putting a task due in on the 13th on hold in favour of one due in on the 19th when I could turn the PC off in an hour and not turn it on until the 14th !!! I'm not sure if I'm getting my meaning across but basically the scheduling of tasks is bonkers and some of my tasks risk not going through on time because boinc has to look clever and run more stuff at once than it is really running. I have tried setting the "switch between applications every "x" minute" to 1000 minutes but still no luck so that setting elludes me and frankly I'm just fed up.

Am i missing something or does this aspect of the software really need looking at please. I run boinc on a whole range on computers, I don't think the computing power is being efficiently used, if tasks miss deadlines then I'm guessing all of those hours work have been a waste of timer, you never know when a PC will be truned off and how long it may remain off, so please get the tasks done with the least time left first and don't keep swapping between tasks because you may loose them both

---

you never know when a PC will be turned off and how long it may remain off

And so how is BOINC supposed to know this? Through some Magic 8 ball encryption, of some sort? Maybe some integration into your calendar, and that you then will put in there when you are going to turn the computer off and back on?

Because come on, really now, if you don't know when you're going to turn computers on and off, then how is the software you run on it supposed to know?
Besides, that is why the redundancy factor is here, where tasks get sent to multiple computers until they're done. If you find that's a waste, then do go make a better version.

Or demand from all projects that they do away with finicky things like deadlines, so you can go turn your computer off when you feel like it, BOINC and work be damned.

---

my problem is where I'm part way through a task with a 12/02/2012 deadline and boinc decides it will stop doing that task and instead start on a task that is due in on the 18/02/2012 (I have this situation on an old XP1600 pc that I'm running just boinc on right now). What is the logic of that ?

I need another 2.5 hours to complete the task due in first, if that task is set aside in favour of a new task that is due in a week later the more urgent task could get overlooked if I happen to turn the PC off for a couple of days, why start 2 tasks at once ? it increases the risk of both being lost as opposed to one getting sent in just on time and one that was never started getting dropped.

I understand that there is built in redundancy to the overall "grid" but this seems to be all the more needed to cope with this way of working, why can't one task at a time be started ? there is no logic in starting two at once and running the risk of getting neither done. even my main quad core PC is doing this it just does not make sense, one task at a time, at the end of the day it aint gonna go any faster.

---

basically the scheduling of tasks is bonkers

It does indeed seem very strange sometimes.

Rather than fight it, and just be told a million times that you are wrong and that you shouldn't micromanage etc etc, why not just set a zero cache and tiny 'extra work' time (e.g. 0 and 0.01) so that your machine gets just one WU per core that runs to completion and then requests new work? This assumes that you have a 24/7 net connection of course!

Cheers,

Al.

yes I'm trying that too now but really caching work to me means having work stored ready and doing it when the last task is finished. I do indeed have a couple of computers that I can't guarantee to have connected all the time and one I only connect when it needs more work so if it has a WU ready so that when the current one finishes is has more work immediately it won't sit idle until i notice it's done.

---

Just increase the 'Computer is connected to the Internet about every x days' setting, while decreasing the 'Maintain enough work for an additional y days' setting,
Boinc will then try and do and report the work x days before deadline, and will cache x + y days of work,
(Note: this is for Boinc 6.12.x and earlier; Note 2: on some Projects the wording has changed in preparation for Boinc 7)

Claggy

sounds mighty complicated for the average user. i may be able to wade my way through this logic but most people simply won't.

I now remember why I set some time of caching - it was because some projects have trouble connecting to the project servers so i had a situation where work had finished and new work was not being downloaded. Either way i have trouble.

What is the "switch between projects every "x" minutes" setting for ? I can't see that it has any effect at all. while we are on that little set of settings what about the use "x"% of cores ? I mean what the ???? why not just specify number of cores ? I mean when i have 4 cores it looks silly but ok I know that it is 25% per core, but when a user has a 7 core processor its not so simple. I really think some of those settings need making better and simpler

---

if I have set 0.2 days to be cached why were 2 WU's downloaded ? an 11 hour and over 30 hours (possibly 60).......

---

Given the variety of client, network and server outages that can occur on both scheduled and unplanned bases, the artificial intelligence built into scheduling tasks can make boinc operate way below optimum.

And on modern multi-processor systems on which multiple projects can be run simultaneously, settings such as "switch between applications" can more often be deleterious than helpful.

It would be valuable to have a global switch to turn off the AI in favor of simpler prioritization by deadline and receipt within fractional project allocation.

Is there a document or posting that summarizes what goals and features the upcoming boinc release strives for?

---

Win 10-64 Pro on: Dual Xeon Quad E5472s 3.0GHz w/128GB DDR2 Main Memory + ATI FirePro W5000 GPU; Quad+HT i7-860 2.8GHz 8GB + ATI FirePro V4800 GPU; Quad Q6600-775 2.4GHz 4GB + ATI FirePro V4800 GPU; + 3 laptops

I am ever more exasperated ! I have boinc on my work PC, this is a celeron 2.8 GHz and does not have to do any heavy work just office stuff and printing out the odd cad drawing. It runs on a very fixed weekly schedule.

yet I still have to micromanage things because even with a known working pattern the stupid damn thing ditched a task that had a deadline that evening yet only 8 minutes to go to work on another task that had a week to go, so i had to halt that to force it to finish the other task first. This is bonkers ! the task management is basically not fit for purpose !

the only reason I can accept 2 tasks being started at once is when much more ram is allowed to be used when in "standby" allowing the PC to work on tasks with more memory requirements while I'm away and switching back to a task with a lower ram usage while I'm using it. To do this however there needs to be a much better coordination between tasks and possibly the boinc manager being able to know properties of each task in advance and make proper intelligent decisions.

Sorry to upset a proud programmer but facts are facts and progress is made by accepting that things can be improved.

---

yet I still have to micromanage things

Says who? Who is forcing you to do these things? Why not just let it go, and let that task or those tasks go past the deadline? How else is BOINC ever going to learn for itself when you are making all the decisions? Why can't you let go?

I haven't checked on my running BOINC for a week now. It's probably doing something, since the CPU is hotter than during idle hours.

Yet if you think something is broken, add the appropriate debug flags (<rr_simulation/> and <cpu_sched_debug/> in this case) and send a report in to the BOINC alpha email list (registration required). Or go get the source code, fix whatever you think is broken, compile a client, test it out and tell the developers about it.

Until that time, and until you can let go of doing all the scheduling your self, you ain't got a leg to stand on.

---

Is there a document or posting that summarizes what goals and features the upcoming boinc release strives for?

Sorry, my FAQ isn't ready for mainstream yet.

In the mean time, parts of http://boinc.berkeley.edu/trac/wiki/ClientSchedOctTen and [url=http://boinc.berkeley.edu/trac/wiki/CreditNew[/url] will be included, as well as full(er) control options for VMs. Oh, and the new simple GUI, of course, http://boinc.berkeley.edu/wiki/New_Simple_View_Design.

---

you never know when a PC will be turned off and how long it may remain off

And so how is BOINC supposed to know this?

I basically agree with the original post as I posted something a bit more politically correct. So how?

1. Calculate the average time that the computer is allowed to run BOINC by using a cumulative average over each period of run time, and compound that average over all successive periods of run time.
2. Pay attention to the "keep enough work for X days" field in the BOINC options.
3. Download only enough WUs to satisfy the "keep enough work for X days" field based on the average run time allowed to BOINC as calculated above even if it is only one WU.
4. Prioritize by due by date.

---

Just increase the 'Computer is connected to the Internet about every x days' setting, while decreasing the 'Maintain enough work for an additional y days' setting,
Boinc will then try and do and report the work x days before deadline, and will cache x + y days of work,
(Note: this is for Boinc 6.12.x and earlier; Note 2: on some Projects the wording has changed in preparation for Boinc 7)

Claggy

BOINC pays no attention to the 'Maintain enough work for an additional y days' setting, See http://boinc.berkeley.edu/dev/forum_thread.php?id=7279 and http://boinc.berkeley.edu/dev/forum_thread.php?id=7280

---

yet I still have to micromanage things

Says who? Who is forcing you to do these things? Why not just let it go, and let that task or those tasks go past the deadline? How else is BOINC ever going to learn for itself when you are making all the decisions? Why can't you let go?

I haven't checked on my running BOINC for a week now. It's probably doing something, since the CPU is hotter than during idle hours.

Yet if you think something is broken, add the appropriate debug flags (<rr_simulation/> and <cpu_sched_debug/> in this case) and send a report in to the BOINC alpha email list (registration required). Or go get the source code, fix whatever you think is broken, compile a client, test it out and tell the developers about it.

Until that time, and until you can let go of doing all the scheduling your self, you ain't got a leg to stand on.

My my, sounds incendiary to me. People like the original poster at least care enough about the volunteer aspect of BOINC to post items like this. While the over-all effort might not miss us, we do, of our own volition, contribute our processing power to the various projects, and then we should, perhaps, learn to program, and then program our way out of things like this? Anything else you would like us to do??

Personally, I've let go of scheduling. BOINC does not learn.

There's an old adage - "You'll catch more bears with honey."

---

does boinc feed back information about what is going on ? tasks being missed and what the settings are. If this information was sent back then the development team would have the facts and can act on them.

I've not actually found the "switch projects every "x" minutes" to have any effect at all on anything and worse it can't be simply disabled.boinc seems bent on making.

To be honest having certain log files demanded of me to prove my case is not helpful, I've described the problem and whoever wrote the program knows what they did. Please just consider developing it further.

---

To be honest having certain log files demanded of me to prove my case is not helpful...

It's exactly one log file (stdoutdae.txt).

I've described the problem and whoever wrote the program knows what they did...

They think they know, but if the program doesn't behave as expected, they are dependent on the users to provide debugging information if they want to develop it further.

Gruß,
Gundolf

ok so where do I find this file and who do i send it to ?

I think basically it is not fault/bug but just that the software needs to be improved to be more advanced at project management.

i get the impression that it is designed to give the impression you are working on all of your projects at once so starts more than one WU at once. I am happy to know that over the period of a month my pc has done it's allotted amount of work and that allows for one Wu at a time to be started.

It would be nice if boinc can in the case of boinc being set to use much more ram when the pc is idle than when in use starts two WU, one that needs more ram and one that needs less, but in this case it must manage the situation and I don't think it is currently built to do that

---

ok so where do I find this file and who do i send it to ?

I think basically it is not fault/bug but just that the software needs to be improved to be more advanced at project management.

i get the impression that it is designed to give the impression you are working on all of your projects at once so starts more than one WU at once. I am happy to know that over the period of a month my pc has done it's allotted amount of work and that allows for one Wu at a time to be started.

It would be nice if boinc can in the case of boinc being set to use much more ram when the pc is idle than when in use starts two WU, one that needs more ram and one that needs less, but in this case it must manage the situation and I don't think it is currently built to do that

---

I run 3 projects on BOINC - Einstein, Milkyway, and SETI. Yet my PC will be sitting here for over 9 hours with no WU to process. Looking at the Event Log a request to Einstein says I don't want any more tasks despite the fact I didn't have any, Milkyway said my PC was full up with tasks, and SETI doesn't have enough tasks to send anyway.

There is something wrong with the scheduling process.

I just find it mildly frustrating to be in a position to volunteer time and money yet have it apparently spurned as this sort of problem occurs not infrequently.

There must be a lot of wasted power out there because of this type of problem. Hopefully it may sometime perhaps be fixed in the near future.