second day -

Wed 12 Feb 2014 10:17:07 PM EET | SETI@home | update requested by user
Wed 12 Feb 2014 10:17:12 PM EET | SETI@home | Sending scheduler request: Requested by user.
Wed 12 Feb 2014 10:17:12 PM EET | SETI@home | Not requesting tasks: don't need
Wed 12 Feb 2014 10:17:15 PM EET | SETI@home | Scheduler request completed


- wtf is this? why "don't need"???
i _need_ in my computer new tasks! but on "update project" ( i do not know another way to manually startd to get new tasks for project) i got that.
app_info.xml is ap for nvidia and mb for cuda - in any way there no "no new tasks for your configuration" or so on....
btw, why there, in client, not insorporate manual commands like "get 10 new tasks for GPU", CPU and so on?

In your BOINC Computing Preferences, http://setiathome.berkeley.edu/prefs.php?subset=global, what is "Maintain enough tasks to keep busy for at least" set to?

Cheers.

Maintain enough tasks to keep busy for at least
(max 10 days). 0.1 days
... and up to an additional 0.5 days


but i not get even 1 task at all. not have any task, and not get any new.
anyway, ok, i change that positions to 3 days and 5 days....

I can see 1 tasks there, but yes increasing that setting should get you more.

Cheers.

no,no,that my account works ok ( thank to god). i not get new tasks on at least two computers on another my account:

http://setiathome.berkeley.edu/show_host_detail.php?hostid=7210225

btw, changed min buffer values, including in global_prefs_override, restart client, do ./boinccmd --project setiathome-bla-bla-bla update
- get the same: not need new tasks :-O
i got milkyway, and another projects tasks without problem, but not seti :-O

no,no,that my account works ok ( thank to god). i not get new tasks on at least two computers on another my account:

http://setiathome.berkeley.edu/show_host_detail.php?hostid=7210225

btw, changed min buffer values, including in global_prefs_override, restart client, do ./boinccmd --project setiathome-bla-bla-bla update
- get the same: not need new tasks :-O
i got milkyway, and another projects tasks without problem, but not seti :-O

Oh this rig is on another account (you should've stated that to start with) so it's likely that your other projects are taking up all the time set aside so actually setting them to no new tasks for a while will eventually get you new SETI tasks again.

Cheers.

but share resources =100 for all projects, as so, as i understand, there must be got tasks from all projects in the same proportion, not?
not "i get 50 wu of milkyway, and dont need any wu from another 5 projects!" ? :)

I will let someone who knows more about those other projects of yours reply then as I have SETI set to 100 and my backup projects set to 0 (or a close to 0 as I can) so I only get work from them when I actually run out of SETI work.

Cheers.

ok, thank you for idea....

but share resources =100 for all projects, as so, as i understand, there must be got tasks from all projects in the same proportion, not?
not "i get 50 wu of milkyway, and dont need any wu from another 5 projects!" ? :)

You'll get work from all your projects in that share long term, not short term, Boinc will ripple through the projects asking for work, whichever project has the highest priority will get asked first,
If that project, ie Milkyway has work available, you cache will get filled, now there is no need to ask for work from Seti as the cache is Full,
as you crunch the priority will slowly switch back to Seti (or whatever other project that host is attached to), eventually those projects will get their turn to ask first.

Claggy

Resources are not shared on an hourly or daily basis.

They are shared on a weekly to monthly basis. For several days
Boinc will allocate resources to one project, then when some counter reaches
the right point will realocate resources to a different project.

Managing resources to a daily basis means you will have to babysit them on a daily basis.

Resources are not shared on an hourly or daily basis.
They are shared on a weekly to monthly basis.

This idea is extremely wrong. BOINC authors (or David Anderson) missed one important thing - a lot of projects have work only sporadic. In newest versions of BOINC manager such projects are not sharing resources, they are SNATCHING resources. One from other.
Or Project 1 (P1) is blocking Project 2 (P2), when P1 finally got work. So, You can set resource share 1000 for P2, and 100 for P1, BOINC manager WILL NOT hunt tasks for P2. The worst thing here - if I will try to get work for P2 manually (suspending P1), P1 may start run in high priority mode immediately. When P1=CPDN (with 1 year long deadline and estimated task processing time for 1 month) and I see such situation, I want to hit my head with my keyboard. :-)
The behavior of BOINC manager is more and more unexplainable (for a single mortal). In reality resource share, additional work buffer are no more working.
Finally - we have second Credit New.

Managing resources to a daily basis means you will have to babysit them on a daily basis.

Your arguments, please. ;-) The worst thing is - we have no possibility to choose between 2 alternatives:
1. to use super-AI of BOINC manager;
2. turn it off.
P.S. Sorry, for brutal tone, but I am too much discouraged.

---

If you really don't have the patience to wait while BOINC evens everything out (it does eventually, honestly), set

<rec_half_life_days>X</rec_half_life_days>

to one day (or less), as described in client configuration

Thank You very much, of course, but I need time to decode. :-)
Edit:
<rec_half_life_days>X</rec_half_life_days>
A project's scheduling priority is determined by its estimated credit in the last X days. Default is 10; set it larger if you run long high-priority jobs.
My English is too poor to understand THIS.
Question.
If I want to use resource share "AS IS", I need to set X, for example, to 0.01. Yes?
Next question.
How to set highest priority for Project N?

---

Resources are not shared on an hourly or daily basis.
They are shared on a weekly to monthly basis.

This idea is extremely wrong. BOINC authors (or David Anderson) missed one important thing - a lot of projects have work only sporadic. In newest versions of BOINC manager such projects are not sharing resources, they are SNATCHING resources. One from other.

I personally don't think this is BOINC's (or David Anderson's) fault. It is the fault of the projects for giving unnecessarily short deadlines. Case in point, I belonged to a project called Superlink@Technion which gave 3 day deadlines. This pretty much assured that their workunits would always get preferential CPU time over all other projects as BOINC didn't want to miss their deadlines. There's no reason why the deadlines had to be this short. I also used to crunch fro CPDN but their really long deadlines meant that they would be pushed to the back of the line until it was very near the deadline. Again, this is unnecessary.

If Project Administrators were more realistic with their deadlines, these things wouldn't happen. The only way BOINC could fix this problem is to dictate that every workunit on every project have the same fixed deadline. I suspect were that to happen, people would be complaining and moaning that David Anderson overstepped his authority again.

It's really a no-win situation, but you need to blame the right group. Not everything is BOINC's fault.

> personally don't think this is BOINC's (or David Anderson's) fault. It is the >fault of the projects for giving unnecessarily short deadlines.

ok, that's not his fault, this is an foult because someone not thinking about that situation - but who do not let him incorporate "manual control" in boinc?
like as
boinccmd --project setiathome GET 10 WU

?
and also possibility to switching manually between tasks also will be good thing imho. now we can only "suspend" and "resume"....

---

?
and also possibility to switching manually between tasks also will be good thing imho. now we can only "suspend" and "resume"....

Because you have to remember that BOINC is designed with "set it and forget it" in mind, for the casual user. It is only power users that want to tweak and force BOINC to work the way they want it to and not the way it was designed.

Personally, it amazes me how much people want to micromanage BOINC instead of just leaving it alone. BOINC is designed, rather well, to keep the cache full and to not miss deadlines. The debt system is often difficult for people to understand, but like all systems if you leave them alone they do their job.

Because you have to remember that BOINC is designed with "set it and forget it" in mind, for the casual user.

Unfortunately, You have to remember something too. When BOINC started, official point of view was: "BOINC never will be "fire and forget" system".

It is only power users that want to tweak and force BOINC to work the way they want it to and not the way it was designed.

If BOINC was designed for "do not disturb me" mode only then explain, please - why we have at all buttons like "Suspend", "No new tasks" etc. :-)

Personally, it amazes me how much people want to micromanage BOINC instead of just leaving it alone.

Not micromanage, but MANAGE. If I want to give some instructions for BOINC manager (forward - BM), I really know, what I do and what I want.
You do not understand people like me? The explanation of this may be - all projects, supported by You, are for You on the same level of priority or importance + all of them have constant flow of tasks. Am I right?
If so, my point of view is extremely different.
Trying to explain on my project list (seen on the picture below).
Level 1 - CERN. LHC - the project of highest priority (unfortunately, never has a constant flow of tasks).
So, every my computer must hunt for tasks from LHC and, if the tasks are finally captured, BM must forget everything and process only LHC.
Level 2 - astronomy and climatology, but in proportion 2:1.
So, I want to donate my CPU time every day and permanent for SETI, Einstein and CPDN in proportion 2:2:1. If CPDN does not give a constant flow of tasks, this is the problem of CPDN, but not the problem of SETI or Einstein. To hell, I want to see all 3 projects always working together and exactly in proportion, I have set. I do not want to miss interesting tasks from astronomic projects.
In earlier versions of BM the game was simple, by setting resource share: 1000 - LHC, 200 - SETI and Einstein, 100 - CPDN. All this worked excellent.
Today I see extremely different situation: if CPDN (a project with huge tasks and year long deadlines) got work after some longer break, LHC is blocked at all - BM is not asking LHC for work, if I ask manually, I see "Don't need" or "Project has no highest priority". The result is - my computers are missing batches of tasks from LHC day after day, month after month. So, I am sitting powerless and thinking: "It is time to hit this damn computer to the wall!"

BOINC is designed, rather well, to keep the cache full and to not miss deadlines.

Unfortunately, this is no more working. An example from my projects:
1) when CPDN captured work, the BM is retargeting to CPDN all resources, trying to compensate for CPDN the lost time;
2) the Einsteins are waiting day after day, until high priority mode; because BM, as usual, does not recognize correct the estimated time for processing, at least a part of Einsteins are missing their deadlines.
If I forgot to watch the process and abort the tasks, which will be too late, I am wasting my CPU time.
P.S. I got a good suggestion, so I will try to make some intervention into those XMLs. Maybe, this will help, at least partially. But anyway, such features are hidden from beginners or casual users - this is very bad situation.
The most important configuration tools must be seen and easy to use for everyone.

---

Because you have to remember that BOINC is designed with "set it and forget it" in mind, for the casual user.

Unfortunately, You have to remember something too. When BOINC started, official point of view was: "BOINC never will be "fire and forget" system".

I've been using BOINC since the beginning and I've never heard them say that BOINC will never be a set-it-and-forget-it system.

It is only power users that want to tweak and force BOINC to work the way they want it to and not the way it was designed.

If BOINC was designed for "do not disturb me" mode only then explain, please - why we have at all buttons like "Suspend", "No new tasks" etc. :-)

You're being (purposefully?) obtuse. Of course there's buttons to do things in BOINC, but merely having buttons doesn't correlate automatically into "it is not designed as a set-it-and-forget-it" system.

Personally, it amazes me how much people want to micromanage BOINC instead of just leaving it alone.

Not micromanage, but MANAGE. If I want to give some instructions for BOINC manager (forward - BM), I really know, what I do and what I want.

Any attempt to circumvent the design because it makes no sense to you, and any attempt to force BOINC into behaving how you think it should behave is by definition micromanaging. I'm sure you know what you want and how you want it to behave, which is why you believe you have to micromanage the system.

You do not understand people like me? The explanation of this may be - all projects, supported by You, are for You on the same level of priority or importance + all of them have constant flow of tasks. Am I right?

No. I have my resources set differently for each project, like you do. The difference between us is that I believe I understand the long-term debt system of BOINC and know that my resource shares will be honored over time if I leave it alone.

If so, my point of view is extremely different.
Trying to explain on my project list (seen on the picture below).
Level 1 - CERN. LHC - the project of highest priority (unfortunately, never has a constant flow of tasks).
So, every my computer must hunt for tasks from LHC and, if the tasks are finally captured, BM must forget everything and process only LHC.
Level 2 - astronomy and climatology, but in proportion 2:1.
So, I want to donate my CPU time every day and permanent for SETI, Einstein and CPDN in proportion 2:2:1. If CPDN does not give a constant flow of tasks, this is the problem of CPDN, but not the problem of SETI or Einstein. To hell, I want to see all 3 projects always working together and exactly in proportion, I have set. I do not want to miss interesting tasks from astronomic projects.
In earlier versions of BM the game was simple, by setting resource share: 1000 - LHC, 200 - SETI and Einstein, 100 - CPDN. All this worked excellent.

There's the problem. You want to see BOINC behaving a specific way and you're going to force it to work that way even if it means interrupting the system's long term and short term debts. Again, this is the very definition of micromanaging. The system largely still works the same, or at least it does for me. However, I haven't screwed up my debt systems by micromanaging, so BOINC behaves exactly as I expect it to in the long term.

Today I see extremely different situation: if CPDN (a project with huge tasks and year long deadlines) got work after some longer break, LHC is blocked at all - BM is not asking LHC for work, if I ask manually, I see "Don't need" or "Project has no highest priority". The result is - my computers are missing batches of tasks from LHC day after day, month after month. So, I am sitting powerless and thinking: "It is time to hit this damn computer to the wall!"

Again, this leads back to the Project Administrator giving unrealistic deadlines to their workunits. The extreme cases like CPDN and Superlink@Technion are some of the outliers that really mess up the debt system.

Then, because the human (you) do not see the behavior that you want, you believe intervention is required to make the behavior better, but in reality you're making it worse. Finally, you have the human emotion of your frustration of wanting to "throw the computer to the wall". You could avoid that frustration if you just let the computer do it's job and not micromanage.

BOINC is designed, rather well, to keep the cache full and to not miss deadlines.

Unfortunately, this is no more working. An example from my projects:

But it is working the way it is designed. The problem is it is not behaving the way you expect it to.


Let me draw a parable for you to illustrate the problem. My current boss is a micromanager. He doesn't believe he's a micromanager, he simply expects us to work a very specific way. I've tried telling him to leave us to do our job and trust us as professional adults to do our jobs correctly. The purpose of management, IMO, is to lead and inspire your people to do better, but to otherwise leave the details to the people doing the job daily. Direct them if they have questions or are stumped. You don't need to tell them how to perform each minute detail of the job.

Fortunately for me, I'm getting a promotion out of my current position and won't be under him much longer. I can only hope my new boss isn't a micromanager.

But the point is, you are much like him. You want things to happen a specific way. You want to tell BOINC how to do it's job. If BOINC doesn't do it the way you want it to happen, you intervene; by definition you micromanage the system. You don't believe you're micromanaging, much like my current boss, you just want things to go a very specific way. Micromanagers often don't realize that they're micromanagers. (Incidentally, according to a recent survey that I cannot find or link to at the moment, micromanagers are frequently some of the worst managers out there - and all of them deny being one).

So again, I'm sure you know what you want to happen on your system. And yes, you can micromanage to get it to do it. What I'm saying is that it is truly unnecessary, and it doesn't need to be precisely as you expect it to be. Time is money, and close enough should be good enough unless high precision is an absolute requirement for the job at hand.

P.S. I got a good suggestion, so I will try to make some intervention into those XMLs. Maybe, this will help, at least partially. But anyway, such features are hidden from beginners or casual users - this is very bad situation.
The most important configuration tools must be seen and easy to use for everyone.

You're assuming everyone wants to micromanage things, and you're assuming everyone wants to spend the amount of time you do into forcing the system to behave how you expect it to behave. Most people don't care because they have other things to worry about.

i agree with metalius.
and i also understand your point - boinc is designed as mass product. and masses is only from far distance looks like as "homo sapiens" - when you get closer, there will be only humans, without sapienism traces, sadly. that is why "microsoft way" is so popular in masses - they do philosophy "users is idiots, we know better what they need!" - with all that choices from one button - "ok", and non existent way to change lot of settings ( most it can do via registry editing).
but i not see, why in boinc development not to involve some more possibilities for that peoples, who want some controls, and using no graphical interface - why do not make commandliine version bit better and with more posiibilities?
i think, it not harm any...

---