I have CP.NET configured to use 38.68% of the resource share on my computer. I also have 6 other projects each of which are set to 10-12%. I leave my computer \"on\" 24/7 and BOINC is set to \"run always\". I have BOINC configured to \"switch every 120 minutes\".

PROBLEM:
-------
CP.NET hasn\'t run on my computer in the past 4 days.

PROBABLE REASON ?
---------------
The new BOINC manager does *not* cycle through my projects and allocate time precisely and FAIRLY based on my settings. It factors in the DUE DATE of individual work units and give it\'s OWN priority to work untis that are closer to their due dates.

My CP.NET due date is one year in the future. PRIME GRID due dates are typically a few *days* in the future. The other projects tend to more reasonable and are a few weeks in the future.

But regardless, I don\'t care about the project\'s due date. When I say give CP.NET 38.68% of my computer I expect it to be running 38.68% of the time.

I feel that the other projects can, and are, hijacking extra resource share by assigning short due dates.

IMO, the BOINC manager is too smart for it\'s good. It should offer the option of ignoring due dates. If a project wants to ignore my results because I\'m a day late, then so be it. I\'ll simply drop the project.

--- bt

It will all work out in the long run, just wait and watch.....Works like that on my computer and I am running an 840ee 3.2 ht with 6 projects.....

---

BOINC Wiki

It will all work out in the long run, just wait and watch.....

Perhaps it will, but that\'s *not* the point.

The BOINC manager should *not* care about arbitrary deadlines set by the individual projects. It should time-slice the CPU on my computer as per my settings. It may have to do a little bit of juggling based on the percentages, but it\'s not hard to work out a simple algorithm.

I should have this choice !! That\'s all I\'m asking.

If one project has set really tight deadlines, the other projects should not have to be sacraficed. That is rewarding bad behaviour.

I have 7 computers running 7 projects, in various configurations. What I may have to do is limit one project per machine so that one project cannot hijack all of my machines ... but that defeats one of the big advantages of BOINC, sharing, and sharing is what it\'s all about.

--- bt

---

I do not think setting a short deadline is bad behavior, as some projects have different requirements. The deadline is the most important item just after the science, You can micromanage if you wish,but to me that is a waste of time.. I have never missed a deadline and thus the science requirement is met and that is what it is about......

This is what affects the preferences

EDF is caused by:

1) A deadline within 24 hours.
2) A deadline within 2 * the connect time.
3) A failure of the Round Robin simulator to finish a result within 90% of its deadline.

A project not requesting work is caused by:
1) A host that is in NWF
2) A project that has enough work on a host that has enough work.
3) A project that has a LTD that is negative enough.

NWF is caused by:
1) A failure of the Round Robin simulator to get a result done within 90% of a deadline if the resource share of the next project to request work from is added to the Round Robin simulation.

Work will always be requested from somewhere, even if that somewhere has a very negative LTD and/or the host is in NWF if there is a CPU that is idle and there is a network connection.
JM7

---

BOINC Wiki

JIm said

It will all work out in the long run, just wait and watch...

In the long run, yes. As a rule of thumb divide each project\'s deadline by its resource share. So if Einstein (14days) has 25% that comes to 8 weeks 92 / 0,25); if predictor (7 days) has 10% (1 / 0.1) that comes to 10 weeks; and so on. Whichever division gives the longest is a ballpark figure for how soon it will come right.

With collections of projects that don\'t include CPDN the balance usually works quite well, in my opinion. After all, on a fine enough timescale the whole point of time-slicing is that someone has 100% for a while, and then a bit later someone else does.

The problem arises for me when CPDN downloads its new 3x-as-long sulphur wu and *cpdn* goes into EDF. That forces all other work to be done first, then CPDN will presumably hog the box for months... OK OK then CPDN will be blocked for months by the long term debt but that is not really what *I* want.

To summarise then, the issue is really one of \"how soon do each of us expect things to come out to the specified ratios\". Being human, opinions will vary. The disagreement is not a technical one but rather it is a question of personal taste.


You say

The BOINC manager should *not* care about arbitrary deadlines set by the individual projects

but in fact the current scheduling was achieved in response to users\' unhappiness with deadlines being missed under the older, simpler, scheduling that offered exactly what you are asking for. The huge majority wanted deadlines to be honoured both for scientific reasons and to preserve credit.

I am sorry you are not getting what you want but in all fairness to JM7 the software is a splendid technical attempt to reconcile many users\' conflicting tastes.

So even tho it does not do quite what I\'d like, I am happy to accept it as an honest and technically brilliant compromise between differing tastes.

---

[The huge majority wanted deadlines to be honoured both for scientific reasons and to preserve credit.]

I\'m not disputing that - although I would argue that it\'s probably the vast majority of people \"who complained\" wanted deadlines [grin]. I would like to have the option of going with simple time-slicing or the complicated scheduling. That would be as simple as a radio button on the preferences pages (and a new boinc manager)

[The deadline is the most important item just after the science]

Yikes. When did finding the next prime number BY TOMMORROW, become that important ? I will reiterate: the projects decide their own deadlines and the deadlines appear to be arbitrary.

If my machine can\'t keep 7 projects happy (3.5 hours each per day) then I shouldn\'t be running 7 projects.

What is now WORSE, is the fact that my machine is over comitted with 3 projects and refuses to even download a single WU from the other 4 projects. There\'s a problem with BOINC if it can\'t have at least one WU from each project (unless of course the project is down).

BOINC should not download 15 WU\'s at a time for one project and then sit back and say \"gee, I have enough work for 3 days from this one project, so I won\'t even bother getting any more work from the others.\" The best I can do is manually abort WU\'s if I think there are too many for one project. That will hoepfully force BOINC to at least consider getting from the other projects.

--- bt

On the BOINC site, which is where this matter should be aired, there is advice to suspend SETI while they are having problems.
Apparently, if BOINC thinks there is a SETI wu on the way, it keeps waiting for it and ignores the other projects. Even when the machine is only attached to two projects.

And, while cp deadline is one year, it\'s not set in concrete.

---[Apparently, if BOINC thinks there is a SETI WU on the way, it keeps waiting for it and ignores the other projects]

BOINC is trying to be too smart for it\'s own good. The operative word here is \"trying\" because it is not succeeding, and the more they tinker with their complicated algorithm, the worse it gets, IMO.

What I\'m now forced to do is have only one project on each machine. This fundamentally defeats the primary purpose of BOINC, but that\'s the way it goes, because BOINC can\'t divide the WORK up in a fair manner.

Do you guys remember the SETI-QUEUE add-on for the classic SETI ? IMO, it was the prefect solution to most of the problems with classic SETI. CP.NET doesn\'t really have these problems because it has enough work queued up for many months.

---[On the BOINC site, which is where this matter should be aired...]

Problem is, there isn\'t a Q&A on the BOINC site. You have to express your ideas on the SETI site instead, and there are a few unsympathetic \"characters\" who simply dismiss all constructive criticism as being \"anti-progress\" ... unlike the people who are responding on this site (thx).

--- bt

BOINC Q&A is <a href=\"http://boinc.berkeley.edu/dev/\"> here.</a>

I don\'t remember SETI at all, because I\'m a cpdn person. :)

BOINC Q&A is here.

Thx. I didn\'t know this existed.

Anyway, you can read my post over there if you like
http://boinc.berkeley.edu/dev/forum_thread.php?id=268

I actually did get some meaningful discussion started with one guy (Bill Michael), but like I said earlier, within minutes I had a couple of unsympathetic clowns bashing me (JimK and Michael Roycraft).

It\'s hardly worth the effort sometimes.

--- bt

I\'m afraid you view of how BOINC should work is a bit academic.
It has evolved to where it is for various reasons, most of which I don\'t know about.
As you are a long time user here, you will have seen this evolution. Most of the other long time users seem happy with it, as long as the bugs get fixed.
As BOINC mostly does what we want at last, we have no inclination to support change back to where it was earlier in the year. Although it would be good if BOINC was better at estimating completion time. But this mostly bugs new users.

Sorry to rain on your parade, as they say. Whoever \'they\' are in this case.

I\'m afraid your view of how BOINC should work is a bit academic.

I fully understand, even though I disagree with how the scheduler works.

Anyway ... I have solved MY scheduling problems, for the most part. Each of my 7 computers now only runs ONE project at time.

Every week, or so, I\'ll switch them all around - with the exception of CP.NET because it takes many months to finish a WU.

Speaking of which, I only run CP.NET on one machine anyway. All of my other machines were giving \"-5\" errors. After 3 such errors on a machine, I simply dropped CP.NET from the machine.

Interestingly, I have never had a \"-5\" error on this computer. It\'s a Pentium M 1.4 GHz laptop. All of my other machines are AMD (Athlon, Athlon XP, and Athlon 64).

But that\'s another topic.

--- bt

---

Speaking of which, I only run CP.NET on one machine anyway. All of my other machines were giving \"-5\" errors. After 3 such errors on a machine, I simply dropped CP.NET from the machine.

Interestingly, I have never had a \"-5\" error on this computer. It\'s a Pentium M 1.4 GHz laptop. All of my other machines are AMD (Athlon, Athlon XP, and Athlon 64).

But that\'s another topic.

It is, and it\'s a strange one. If you were to start a new thread, it would be relevant to say whether that was with the current version of the application. And what common factors linked the machines beyond the chip manufacturer. But since you have a one machine, one project, policy I imagine it\'s academic now.

---

But since you have a one machine, one project, policy I imagine it\'s academic now.

Exactly, but thx for your interest.

--- bt

---

I am running BOINC on 3 machines and getting credit for
one machine. So please explain how the climate for the
next century will be modelled but the model cannot
merge the statistics for my 3 machines?

---

Why should a single model merge the results from 3 computers?
Each computer will run a model with slightly different set of values for it\'s parameters, thus producing 3 different end results.
The researchers will be able to see what effect the varying of different parameters has on the model.

As for only getting credit for one machine, perhaps the other models haven\'t yet reached the 4th of December, which is when a small amount of data is sent back. And then it may be up to 24 hours before the credit program makes it\'s once a day run through the results to allocate credits for the work returned.

---

Backups: Here

Your credits often take 24 hours after model trickles to be added to your account, sometimes more. This weekend credits will take longer because the servers will be shutting down for maintenance in approx 5 hours from now, until Sunday or Monday.

Credits don\'t get lost. You always receive them in the end.

---

Cpdn news