Have an interesting one.
Last week, I decided to re-attach E@H and P@H.
Both were set to 10%, and seti was still at 100%.
The first thing that happened, was that E@H downloaded a couple of WUs, and P@H downloaded a dozen or so WUs.
It immediately went into compute critical first, because it still had all of the Seti's that I had before I re-attached the two E and P@H.
Of course, it calculated the E@H and P@H as the most critical.
So it went about crunching them.
I still didn't understand completely why E@H and P@H got priority.
Given that I had already a full plate of Seti, THEN added E@H and P@H. It should have seen that Seti was over-committed and crunched it out a little first.
OR, It shouldn't have downloaded that many E@H and P@H, since I was already at the limit for Seti.
Either way, My assumptions on what should be done when must have been wrong, because it started P@H and E@H, and never touched Seti.
Now, to my current dilema.
It finished crunching and downloading and crunching and downloading E@H and P@H all weekend.
Just a day or so ago, it started doing Seti again.
Doing then real good, real fast, as usual.
But one new issue now.
It won't download any new P@H or E@H.
BOINC is in over-committed mode for Seti, and has been for a couple of days now, and won't ask for new work from P@H and E@H.
So somehow, if you have a lot of seti, and only a couple of E@H and P@H, my P@H ALWAYS gets prioritized over seti, regardless of dates.
Same for E@H.
This was the reason several months ago that I quit trying to do cross project.
Several months ago, same thing. It kept going on a big roller coaster for several weeks, and never got into a steady rhythm of doing an E@H, then a few P@H, then a seti or two. It just got stuck in emptying out P@H, and when it finished, it went to E@H, emptied it, then to Seti, and never downloaded another P@H or E@H.
I left one computer running like this for over one month, and never got another P@H or E@H on it.
Now, I just shut down boinc and restarted it so that I could get the messages, and this is what it shows:
7/26/2005 11:50:22 AM||Starting BOINC client version 4.45 for windows_intelx86
7/26/2005 11:50:22 AM||Data directory: C:\Program Files\BOINC
7/26/2005 11:50:22 AM|Einstein@Home|Computer ID: 356898; location: home; project prefs: default
7/26/2005 11:50:22 AM|ProteinPredictorAtHome|Computer ID: 139083; location: home; project prefs: default
7/26/2005 11:50:22 AM|SETI@home|Computer ID: 1197998; location: home; project prefs: default
7/26/2005 11:50:22 AM||General prefs: from SETI@home (last modified 2005-06-03 08:17:20)
7/26/2005 11:50:22 AM||General prefs: no separate prefs for home; using your defaults
7/26/2005 11:50:22 AM||Remote control not allowed; using loopback address
7/26/2005 11:50:22 AM|SETI@home|Resuming computation for result 07fe05aa.21211.23328.853396.53_6 using setiathome version 4.18
7/26/2005 11:50:22 AM||Using earliest-deadline-first scheduling because computer is overcommitted.

I have 26 Setis on this computer, and have completed 16 and downloaded 16 replacements since my last E@H or P@H was downloaded.
It finishes a seti, then downloads a new one, like clock work, however, it doesn't do anything else anymore for P@H or E@H.

True, I have 10% E@H and 10% P@H, and 100% Seti. (equates to 8.333, 8.333, 83.333). And being on 100% of the time, it seems that it should have done 8% work on a E@H last night through today, but it hasn't done 1.

I'm not bitchin' complainin', just stating the state of the system as it is running now, and did yesterday also, and the weekend before, and....

The only way I have found that I can get a E@H is to suspend Seti, then it downloads 1 E@H (note, P@H says it has no work that it can give, it has work but committed to other platforms, why I can't get one, I don't know, not on a limit, must be another reason, maybe it wants more intel compiles than athlon).

Anyway, that is where I am today.
I don't know why it won't ask E@H and P@H for more work.
Only when I suspend seti does it ask. This I don't understand.

---

Thanks, and Keep on crunchin'
John Henry KI4JPL
Sevierville TN

I started with nothing,
and I still have some of it left.
<img src="http://www.boincstats.com/stats/banner.php?cpid=989478996ebd8eadba8f0809051cdde2">

I know, replying to myself. I just hope I don't argue with myself.

To add something strange to the mix.
1 - It wouldn't ask for more E@H until I suspended Seti.
2 - When I restarted Seti, it is now running only E@H.

Go figure that one out.
When I had no E@H, and let the system crunch them all out, it wouldn't run seti.
Only when E@H and P@H were exhausted, and wouldn't download any new E@H or P@H as it finished them, would it start seti.
Then, when Seti is crunching, it will download new Setis, and won't even ask to download any new E@H or P@H.
THEN, when I suspend seti, it asked, and downloaded E@H, and will not run seti now, crunching, finishing, E@H, downloading new E@H, not running seti.

Go figure.

I will let it run for a few days and see if the scheduler gets its feet going normal again.

---

Thanks, and Keep on crunchin'
John Henry KI4JPL
Sevierville TN

I started with nothing,
and I still have some of it left.
<img src="http://www.boincstats.com/stats/banner.php?cpid=989478996ebd8eadba8f0809051cdde2">

When I upgraded the last 4 to 4.45, I did not do anything pst looking at my normal "work" display of BOINC View where it shows me the comptuer's wort tabs. I only watched to make sure than none went off line and did not do any work.

Patience was rewarded because the machines sorted themselves out. I see, on occasion when I do peek in that sometimes the machine will not have work from ALL of the projects at times. But, leaving them alone has worked for me.

I admit that it was tense here when one machine finished its CPDN work and took 4-5 days to start a new one ... but, alls end when well ... or something like that ...

---

I would betr a dollar that you have a high connect to time at least over 3, if so that be your problem.....


A request for 0 second of work means one of three things:

The computer has too much work on hand already to meet deadlines reliably.

The project has enough work on hand, and the computer is not desperate for work.

The project has used too much CPU time for its resource share recently and is blocked from requesting more for a while.

In all of these cases, the best thing to do is to leave it alone, and it will fetch more work when it is ready.
"JM7"

---

Semper Eadem
So long Paul, it has been a hell of a ride.

Park your ego's, fire up the computers, Science YES, Credits No.

I have 4.72 installed. The scheduler's just plain weird. It acts pretty much the same as 4.45 did. I just downloaded 2 new lhc units. BOINC stopped the seti unit that was running and started one of the new downloads. The seti unit's deadline is two days earlier. I exited BOINC and restarted it. It started running the two newest LHC units. The seti and lhc units that are over 50% complete and due on the same day are preempted. I had to suspend the new units to get the partially completed ones to run.

Want some advice from an old fool? Don't fiddle with Boinc.
If a work unit really gets within its last 24 hours before deadline, Boinc will crunch it first.

At all other times, no matter how strange things go, just let it go.
Not only do you have the new scheduler still working its ways out on your computer, you also have the new estimate per work unit 'guesser' at work.

So don't suspend anything, don't do anything, just let it run. You're only working against all of the upgrades!

---

I agree. Last time we had a power outage in town, my computer was off for a quite a few hours. Well, when I rebooted, it did all the things that jshenry is saying. After a few days, everything started getting back into its groove. I didn't have to touch a thing. So as others have said, let it crunch, it will sort itself out. Its doing all of the calculating it needs to do so that it knows how much work to request and not request. :-) Take care, and happy crunching.

---

It's all very well saying words to the effect of:

"Just let Boinc do IT'S thing"...

How do i get it to do MY thing?

It's my PC, I pay for it, and the electric to run it.

I know what I want my PC to do.

I want to do nothing but Seti, unless both Seti @ the PC run dry. Then I would like to do somthing else.

It seems that this is not an option at the moment.

I can only guess that supporters of other projects might feel the same.

I will willingly Fold, burp, puke, or what ever, but want to Seti 1st.

I think thats what miffs some users off.

I would like to set it up to do seti 1000% of the time.

But when, and only when, I have no seti work, I want to get work from somewhere else, then do that. As soon as the Other work is done. Try for more Seti. If no seti, got work from somewhere else. Repeat till Seti work forthcoming.

Just my 1p's worth....

---

Foamy is "Lord and Master".
(Oh, + some Classic WUs too.)

Dorsai, the 4.7x version in Alpha are now just learning how they can be taught to recognize the correct length of a work unit for the various projects.

Once that traject has grown, once your PC in the future has run such a client, can the real resource shares be tackled. At this moment Boinc is learning. It's in the 3rd grade. If you want it to be University grade material overnight, that won't be possible.

But give it a couple of months and we'll get there. :)

---

Dorsai,

Ok, you want BOINC to work the way you want it to. And you don't like the way it works now. This is all fine. And you do have an option. Take the source code and modify it, compile it and have it work the way that you want it to.

The probability is that your reaction is going to be something along the lines of "thats not fair". Well, true. SETI@Home Classic worked a particular way, lots of people liked it, some didn't. But, you could not change a thing so you lived with it.

BOINC is no different. With one exception, if you don't like the way it works, you can roll your own. The only other choice is to live with the way it works. The designers have a plan how BOINC is going to work, and you either live with the way the system is designed, or you make your own version.

Yes, you are a volunteer. But, when you volunteer you do not get a veto vote on anything. You say, I volunteer to do SETI work, fine, the tool is BOINC and you use that ... or you make your own ...

No one is trying to blow you off, but there is this attitude that after volunteering that you have any say in the way that the projects are run. Well, the only tool you have is to vote with your feet. But a cold dose of reality says that none of us are all that valuable. Yes, they lose "x" computers you own ... but the change is going to be so insignificant that no one will see you leave. Sorry, but the truth is that none of us are all that important to the production.

---

I've been looking at the various threads on the both Einstein and on Seti about over committing of BOINC, etc. I haven't really researched it, but it seems to me that are having problems also have resource shares that are very lopsided, as is this one. For the most part, we have never really discussed the effects of resource share, although we have been preaching small connect times to death.

Here is what I think is happening. Think about this and tell me if it is a possibility. I think that BOINC downloads all the WU's that it wants. it sees that SETI has the highest resource share, so it starts crunching seti. After a while it realizes that it needs to work on E@H and P@H, not because it is their turn due to resource share, but it starts working on them because their deadlines are fast approaching.

Now, what happens is that E@H and P@H now have a high debt (relative to the resource share); therefore it starts working on Seti and not downloading work from the other two. Then, when it is forced it will download work from the other two.

Questions? Comments? I would like to see if the others who are having similar problems have resource shares set up like this person.

Have an interesting one.
Last week, I decided to re-attach E@H and P@H.
Both were set to 10%, and seti was still at 100%.
The first thing that happened, was that E@H downloaded a couple of WUs, and P@H downloaded a dozen or so WUs.

---

You're explaining almost exactly how the work scheduler works, Jim.

---

Uhmmm... Thank you, Ageless.

But, based upon their description of the problem, does this seem to be a plausible idea? maybe if their resource shares were more equal, perhaps then they would see BOINC work like they are expecting?

Maybe I just didn't go far enough to explain what I was thinking in my last post. My mind gets ahead of my fingers sometimes.

You're explaining almost exactly how the work scheduler works, Jim.

---

What then would be the reason to have different resource shares, if they would only work correctly when set equal?

To me, looking at his problem from a resource share point of view, I think BOINC is working as designed. It just isn't working as he expected.

I just now looked to see what his connect every X was set to, but I couldn't find it. I would venture a guess, though, that it is set high, thus exacerbating his resource share situation.

What then would be the reason to have different resource shares, if they would only work correctly when set equal?

---

And even then, the present day benchmarks aren't estimating good enough, not even for the optimised clients. So the new Boinc versions 4.7x are now being taught how to estimate a work unit's time.

As I said earlier, this isn't done in 2 weeks. Your son or daughter doesn't graduate from 3rd grade to University either. Now, it may not take years for this to teach Boinc how to do it, but months in the least.

---

--O/T--

Ageless, your friends are at it again. I looked down through here and all of your posts had -1. I think I need some friends like yours. LOL

Jim

And even then, the present day benchmarks aren't estimating good enough, not even for the optimised clients. So the new Boinc versions 4.7x are now being taught how to estimate a work unit's time.

As I said earlier, this isn't done in 2 weeks. Your son or daughter doesn't graduate from 3rd grade to University either. Now, it may not take years for this to teach Boinc how to do it, but months in the least.

---

Stop calling me Ageless and maybe they go away. You're attracting bad people. ;)
I'm Jord.

---

[quote]To me, looking at his problem from a resource share point of view, I think BOINC is working as designed. It just isn't working as he expected.

I just now looked to see what his connect every X was set to, but I couldn't find it. I would venture a guess, though, that it is set high, thus exacerbating his resource share situation.

[quote]

I have a pretty lopsided share 85% and 15% for the two projects I run all the time on several of my computers. I set the other projects I run at 5%, but don't run the 85% project on the same computers, so in effect the "other" projects are relatively equal.

I'm thinking it may take too long to select a share that works with "all" projects on one or each computer.

Some math whiz should do an Excel (or whatever) that we could plug our projects into and it would give a "more or less" optimum share for the projects with widely varying times.

If you're willing to change the connect time rather than the ration, there's a VERY easy formula... shortest return time (usually 7 days for E@H or PP@H) divided by number of projects.

Therefore, with 3 projects (including Einstein)

Max connect= 7days/3projects= 2.333 days

---