Hi,

Next to seti i am also running CPDN. In the general perfs the recommended setting of 60 minutes to switch between applications confuses me.

The resource sharing is as follows: 99.9 seti and 0.01 CPDN. I have CPDN mainly as backup project, for when seti is off.

On what way is the time divided now? At this moment my pc is on a day, switchin 24 times between apps. So 12 hours seti is on, and the other 12 hours CPDN. This was not really my intention. Or is it going fine after a time? How is this going? I have no experience with double projects......

---

What version of BOINC are you running?

Something else that you might want to consider. Unless your machine is extremely fast, your CPDN WU is going to go into Earliest Deadline First mode at some point so that it can finish before the deadline.

I know what you are trying to accomplish, but I'm not sure if this is they way you want to accomplish it.

Jim

Hi,

Next to seti i am also running CPDN. In the general perfs the recommended setting of 60 minutes to switch between applications confuses me.

The resource sharing is as follows: 99.9 seti and 0.01 CPDN. I have CPDN mainly as backup project, for when seti is off.

On what way is the time divided now? At this moment my pc is on a day, switchin 24 times between apps. So 12 hours seti is on, and the other 12 hours CPDN. This was not really my intention. Or is it going fine after a time? How is this going? I have no experience with double projects......

---

CPDN is not a good choice for a 'backup' project. It's long runtimes will get on your nerves in that setup eventually. Predictor@Home and SETI@Home are currently the best choices for a backup project. They both have relatively short runtimes, loose deadlines and reasonably reliable work availablity.

With the default settings there is an opportunity for a switch every hour. It may not happen, the scheduler may decide to work on the same project next hour as it did last hour. After a settling in period you should see it running your main project 23 out of 24 hours. Depending on your settings it can take a few days to a few weeks for things to settle in.

---

BOINC WIKI

BOINCing since 2002/12/8

Hi Harry,

I have noticed that Boinc does not seem to like running fractions in resource sharing. During the aftermath of big SETI outage to ease the load I changed settings to SETI=100 and Einstein=150 thinking it run 1 hr on SETI then 1.5Hrs on Einstein. What I was actually getting was that Einstein would run for 2hrs in the first cycle then for 1hr 2nd cycle.

I suspect that what is happening to you but with your percentange mix it will take a long time be it sorts it self out. It would appear that you are running 4.45 which will mean that it will go into panic mode to complete WU in time.

As JKeck as said you might have to choose a different project as backup, one with shorter run times but I still think you'll change your mix to something like 90%/10% to prevent WUs overrunning.

---

.


A person who makes no mistakes, creates nothing.

Hi Harry,

I have noticed that Boinc does not seem to like running fractions in resource sharing. During the aftermath of big SETI outage to ease the load I changed settings to SETI=100 and Einstein=150 thinking it run 1 hr on SETI then 1.5Hrs on Einstein. What I was actually getting was that Einstein would run for 2hrs in the first cycle then for 1hr 2nd cycle.

I suspect that what is happening to you but with your percentange mix it will take a long time be it sorts it self out. It would appear that you are running 4.45 which will mean that it will go into panic mode to complete WU in time.

As JKeck as said you might have to choose a different project as backup, one with shorter run times but I still think you'll change your mix to something like 90%/10% to prevent WUs overrunning.

Luigi, There is a preference setting under General Preferences called "switch between applications", it is set to 60 min by default. This means it'll run 60 then decide whether to continue on that project or to switch to a new one. I have mine set to 180 minutes (3 hours). That way it'll finish a seti, ppah, or lhc wu without taking the time to write the file to memory.

The resource share isn't followed by running a project 1 hour, then the next 1.5 hours. It's followed by keeping track of the CPU run time, and calculating the actual run time.

does this help? If not then a link to the wiki is above my sig.

tony

---

Well, i leave it this way then. It wil settle out because the CPDN wu's have a deadline of one year.

But there is no option that BOINC will run the wu with the nearest deadline first?

---

Tony,

Thanks for the info. I have since switched my resource share to 100 on both projects and they generally switch every hour and I am quite happy to leave at that.

From what I read in the forums it would appear Boinc especially 4.45+ does seem to get it knickers in twist when it has to deal with such lopsided resource shares such as Harry's. Perhaps a warning of problems that might ensue should be said somewhere if they use such allocations.

Regards

---

.


A person who makes no mistakes, creates nothing.

As JKeck as said you might have to choose a different project as backup, one with shorter run times but I still think you'll change your mix to something like 90%/10% to prevent WUs overrunning.

(My emphasis added to the quote.)

A very good note.

At anything less than 10% resource share, any WU for any of the projects just won't make adequate progress for sensible scheduling.

In any case, 90% to your first favoured project is a still good chunk of processing.

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Tony,

Thanks for the info. I have since switched my resource share to 100 on both projects and they generally switch every hour and I am quite happy to leave at that.

From what I read in the forums it would appear Boinc especially 4.45+ does seem to get it knickers in twist when it has to deal with such lopsided resource shares such as Harry's. Perhaps a warning of problems that might ensue should be said somewhere if they use such allocations.

Regards

OK, Remember when reading the forums that some report problems that are more a problem of "Appearance" , "missed Expectations", or a general missunderstanding of what is really supposed to happen. This is more prevalent in those that try to micro manage boinc, than in those that trust it and just let it work. The wider the spread in resource share the harder to follow it by watching, or maybe it's just harder to rationalize. The percentages of the resource share will be followed, but it's not a visible thing.

If you have a setting of 5% Einstien and 95% seti, the new scheduler will crunch one or two einsteins first, this will result in a large negative LTD. The Scheduler will then switch to Seti only for a LONG time. This is where confusion comes in because each machine/host and application works slower/faster than the next, and so WE can't say definitively the timing of what happens to YOU. The negative Einstien LTD will slowly become more positive (even more slowly because this is where the 5% come in, it only applies the 5% move in a positive direction). The seti Ltd will become more negative. Once the einstein LTD becomes Nearly Positive it will request more work and the cycle starts again. There is code in the scheduler that says "i'll get work from ANY project if I'm completely out of work". So, if seti goes down and the work runs out, it will get einstien work even if it's not time yet. This ofcourse will make the Einstein LTD even MORE negative than it was before and you'll be crunching LOTS of seti for even longer than before so that the Resource share is honored.

WE users used to have control over what is crunched and when. People got used to that. This has changed and some users just refuse to give up control to a program. they hate change. People got used to balancing resource share by the old train of thought "ok, I do 1 seti so I need 2 PPah to make it equal,...etc".

Oh heck, I forgot where I was going with this.

tony

I think 10% is al little too much for a backup project, but OK, i will change and try.

Edit: Changed it to 99% seti to 1%CPDN, i will see if this works....

---

I think 10% is al little too much for a backup project, but OK, i will change and try.

If it were me and it was merely a back up project and not really a passion I'd set it like "Seti 100, Project X 2" or something along those lines. Did you know a resource share between projects like 1, 1, 1 is the same as 10, 10, 10, or 100, 100, 100. shares of 50, 1, 1 is the same as 100, 2 , 2.

Harry, just find one you like and stick with it. Watch how it operates for a weeks and see if it's working.

If you have a setting of 5% Einstien and 95% seti, the new scheduler will crunch one or two einsteins first, this will result in a large negative LTD. The Scheduler will then switch to Seti only for a LONG time. This is where confusion comes in because each machine/host and application works slower/faster than the next, and so WE can't say definitively the timing of what happens to YOU. The negative Einstien LTD will slowly become more positive (even more slowly because this is where the 5% come in, it only applies the 5% move in a positive direction). The seti Ltd will become more negative. Once the einstein LTD becomes Nearly Positive it will request more work and the cycle starts again. There is code in the scheduler that says "i'll get work from ANY project if I'm completely out of work". So, if seti goes down and the work runs out, it will get einstien work even if it's not time yet. This ofcourse will make the Einstein LTD even MORE negative than it was before and you'll be crunching LOTS of seti for even longer than before so that the Resource share is honored.

What you describe here is JUST THE WAY to have BOINC working on Seti exclusivly unless there are no downloads, in which case it would turn to the other project and fetch a WU from there.

There are people who want do do just that and I think this is the only way it could work. But the second project really should not be CPDN...

:-)= Greybeard

---

All about BOINC: BOINC-Wiki (by Paul D. Buck)

Tony,

Thanks for the info. I have since switched my resource share to 100 on both projects and they generally switch every hour and I am quite happy to leave at that.

From what I read in the forums it would appear Boinc especially 4.45+ does seem to get it knickers in twist when it has to deal with such lopsided resource shares such as Harry's. Perhaps a warning of problems that might ensue should be said somewhere if they use such allocations.

Regards

OK, Remember when reading the forums that some report problems that are more a problem of "Appearance" , "missed Expectations", or a general missunderstanding of what is really supposed to happen. This is more prevalent in those that try to micro manage boinc, than in those that trust it and just let it work. The wider the spread in resource share the harder to follow it by watching, or maybe it's just harder to rationalize. The percentages of the resource share will be followed, but it's not a visible thing.

If you have a setting of 5% Einstien and 95% seti, the new scheduler will crunch one or two einsteins first, this will result in a large negative LTD. The Scheduler will then switch to Seti only for a LONG time. This is where confusion comes in because each machine/host and application works slower/faster than the next, and so WE can't say definitively the timing of what happens to YOU. The negative Einstien LTD will slowly become more positive (even more slowly because this is where the 5% come in, it only applies the 5% move in a positive direction). The seti Ltd will become more negative. Once the einstein LTD becomes Nearly Positive it will request more work and the cycle starts again. There is code in the scheduler that says "i'll get work from ANY project if I'm completely out of work". So, if seti goes down and the work runs out, it will get einstien work even if it's not time yet. This ofcourse will make the Einstein LTD even MORE negative than it was before and you'll be crunching LOTS of seti for even longer than before so that the Resource share is honored.

WE users used to have control over what is crunched and when. People got used to that. This has changed and some users just refuse to give up control to a program. they hate change. People got used to balancing resource share by the old train of thought "ok, I do 1 seti so I need 2 PPah to make it equal,...etc".

Oh heck, I forgot where I was going with this.

tony

Yeh I see where you going and I agree.

Like Harry I started Einstein as a backup project and planned to run 75% SETI and and 25% Einstein but then I found that on my computer Einstein takes 10.75 hrs to complete which is nearly 3 times a SETI WU and only 7 days to it in I had to rejig my priorities especially as as I wanted to increase my cache from I day to 3 days. As said before I now run 50/50 and am happy. It ran thru the DNS change and outage without running out work so I am not complaining. The managing I do now is to update a project if I see there is a WU waiting to report.

Regards

---

.


A person who makes no mistakes, creates nothing.

Forgot to mention that I run BOINC 4.25 and beleive that this does not have STD and LTD thingie. Been wondering if I should upgrade or should I work on the principle "If it ain't broke, don't try to fix it !"

Regards

---

.


A person who makes no mistakes, creates nothing.

Forgot to mention that I run BOINC 4.25 and beleive that this does not have STD and LTD thingie. Been wondering if I should upgrade or should I work on the principle "If it ain't broke, don't try to fix it !"

Regards

sorry, I made the assumption you were using 4.45. I got this from your earlier post about seeing "a problem" with 4.45. 4.25 is the older version manager. It doesn't use the new scheduler, and therefore doesn't have LTD/STD. If you're happy with it, than stick with it. It was a recommended version and as such doesn't have alot of bugs. If you wanna upgrade then feel free. I'd recommend 4.72, but I hear you'll have to manually report the WU after uploading by clicking on the Project tab, then highlighting the project with WUs ready to report, then clicking Update. I haven't seen this "bug" since i'm on dial up and that's how I have to do it anyway. LOL

The new versions have some nice features, but also the confusing scheduler. You pick

tony

Forgot to mention that I run BOINC 4.25 and beleive that this does not have STD and LTD thingie. Been wondering if I should upgrade or should I work on the principle "If it ain't broke, don't try to fix it !"

Boinc 4.45 (and newer) is much better for those of you who want to stick with the main/backup project paradigm. It does really good job even for backup projects with somehow longer WUs as it will notice that deadline is approaching and will temporarily rise the amount of CPU time spent on that WU and back off after it's done. Hence STD/LTD.

The resource share will be met in a bit longer period of time (I expect in 2 or 3 weeks if you wanat to run Einstein as backup for seti).

---

Metod ...

Thanks Tony and Metod,

I think I'll stick with 4.25 as seems to what I want.

Regards

---

.


A person who makes no mistakes, creates nothing.

The problem is that in a year there are 8760 hours.

So if CPDN is getting 1% of this, it is getting 87 hours per year. A CPDN takes around 680 hours (at least on my machine a P4 3.0 HT). So sooner or later it will go into panic mode and crunch exclusively CPDN in order to meet the deadline.

So as already said, an ideal backup project is one with quick units. This way your resource share can be respected, at least in the new clients.

---

Ach, i will see how CPDN preforms, I think it's allright if it takes a little more time, and the wu's are only expiring after a year.......

I will try the preformance!

Edit: and it does not take so long at my athlon 64, because i am running CPDN for 1,5 days now, and i already have gotten credit! How is that possible?

---

...i am running CPDN for 1,5 days now, and i already have gotten credit! How is that possible?

CPDN gives credit after every 1.5% of the WU done, AFAIK.

But then - - this would mean that if you have had 1.5% in 1.5 days, it would take 100 days only to complete. And that at a rate of 1:999... Somewhat queer indeed.
You better have a close look at the behaviour in the next days.

:-)= Greybeard

---

All about BOINC: BOINC-Wiki (by Paul D. Buck)