It started about a week ago, BOINC would Report completed tasks to SETI but would not fetch new work.

1/24/2010 10:58:13 AM SETI@home Sending scheduler request: Requested by user.
1/24/2010 10:58:13 AM SETI@home Reporting 1 completed tasks, not requesting new tasks
1/24/2010 10:58:18 AM SETI@home Scheduler request completed

I updated BOINC to the latest Rev with out success.

I tried turning the fetch function off then on but no change.

I have now run out of work for SETI.

My preferences are set to 10 day work buffer with a connect each day.

I am still getting work from Milkyway but it is limited to only 6 work units with no regard for processing time.

Whats gone wrong?

Why won't BOINC fetch work from SETI?

---

BOINC is not asking for new SETI work, so that means the scheduler believes you have met your resource share or are owing debt to your other project(s), or your computer is not on long enough throughout the course of the week, so "10 days" worth of 8 hours of processing (I'm just guessing here) along with Milkyway won't amount to much.

What is your resource share for Milkyway and SETI?

---

I am still getting work from Milkyway but it is limited to only 6 work units with no regard for processing time.

Whats gone wrong?

Why won't BOINC fetch work from SETI?

BOINC has a long list of rules for when it should and should not fetch work, but for this discussion, the two most important are:

1) Honor resource share

2) Don't miss deadlines

... and at times, these are in conflict. BOINC will concentrate on #2 (making deadlines) and then concentrate on other projects to honor #1 (resource share).

That is why you're asking: most people expect resource share to be honored on a minute-by-minute basis, and it's honored more on a week-to-week (or longer) basis.

To get into the specifics, we'd have to ask about deadlines, resource shares, and the "debt" values internal to BOINC -- and while that's interesting, the answer is going to turn out to be "because BOINC did too much SETI and is now caching up on Milkyway."

---

My computer is running 24/7.

My Resource Share is 50/50 for the two projects.

It is only recently that Milkyway has started sending me work units with any regularity. Most of time it responds with something to the effect that I can not complete the work before the deadline or my machine is not available enough of the time (?).

Today Milkyway responded with 6 work units of about 32 hours each which I think is a gross over estimation of time to completion. The WU that is in process now has an elapsed time plus time to completion of about 28 hours and it has only run a little over 3 hours.

Thanks

---

which makes it off by 1/4 which isnt that big of a deal
if you get overloaded in one project you won't receive work from other projects

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

If Milkyway wasn't sending you work regularly before, then there's a good probability that debt has been worked up for Milkyway while SETI was sending you work in it's place. Now BOINC wants to repay that work by doing "extra" for Milkyway instead of SETI. Eventually it will balance itself out in the long run, assuming both projects continue in a stable manner.

---

My computer is running 24/7.

My Resource Share is 50/50 for the two projects.

It is only recently that Milkyway has started sending me work units with any regularity. Most of time it responds with something to the effect that I can not complete the work before the deadline or my machine is not available enough of the time (?).

Today Milkyway responded with 6 work units of about 32 hours each which I think is a gross over estimation of time to completion. The WU that is in process now has an elapsed time plus time to completion of about 28 hours and it has only run a little over 3 hours.

Thanks

I think you've supplied the answer. Most likely the DCF (Duration Correction Factor) for Milkyway was (and still is) out of wack. This would explain why for a time it refused to get work from MW. While not getting work MW was building up LTD (Long Term Debt). An inflated DCF would also explain the overestimation of the completion times. Both the DCF and the LTD will adjust themselves as you complete work for MW and when the balance is restored work fetch for Seti should return to normal.

---

Boinc V7.2.42
Win7 i5 3.33G 4GB, GTX470

My preferences are set to 10 day work buffer with a connect each day.

Seti is the more patient project too, with deadlines often 6 to 8 weeks away. My experience when managing a large queue across multiple projects is the other projects go red exclamation mark while Seti twiddles it's thumbs and waits (this happened with World Community Grid at least, all their deadlines at the time were like 5 days).

Also, I know the math should wash, but my system seems much happier with a 1 or 2 day buffer of work rather than 10 days. I can also see how a 10 day queue might magnify a problem with DCF so that the bad times appear worse. I also avoid local settings now as that has tripped me up in the past.

Finally, I never got "connect once a day" to work properly. I didn't have time to dive into the sourcecode to see the innards, so I simply conceded that the feature didn't work the way I imagined it would, and put it back to always connected.

My experience with Milky Way is similar. It seems to push the boundries of BOINC's sharing capabilities.

When I do get Milky Way, it can be a large number of tasks with a short dead line. BOINC panics (figuratively speaking) and does nothing but MW at high priority till they are done. This builds up a big debt with the other projects. No more MW is requested while the other projects get back into line, for several days or even a few weeks. When MW is finally requested, I usually get another big list of tasks, and it all starts over.

I have noticed the number of MW tasks recieved at a time is slowly coming down (12 at a time when I started a few months ago, now 8 or 9 at a time). I guess BOINC is "learning" how to handle MW better.

---

My computer is running 24/7.

My Resource Share is 50/50 for the two projects.

It is only recently that Milkyway has started sending me work units with any regularity. Most of time it responds with something to the effect that I can not complete the work before the deadline or my machine is not available enough of the time (?).

Today Milkyway responded with 6 work units of about 32 hours each which I think is a gross over estimation of time to completion. The WU that is in process now has an elapsed time plus time to completion of about 28 hours and it has only run a little over 3 hours.

Thanks

It appears that BOINC is doing exactly what it should be doing.

I agree with OzzFan that the irregular Milkyway work has caused BOINC to think it's overdone SETI -- that time is owed to Milkyway.

The high estimates and number of work units suggests a poor DCF, which will correct itself over time.

... and others have commented about short deadlines.

That adds up to deadline pressure, and BOINC will do its' best to finish what it has before it touches any other project.

The problem that Milkyway is causing for themselves is that your machine will rush to get their work back, the balance of debt will shift in SETI's favor, and you'll have a corresponding time when you have only SETI and BOINC will not fetch Milkyway because it's done too much.

By the way, because you're running two projects, I'd lower the size of your cache. That makes work requests smaller, and eases up on the deadline pressure that causes your complaint. 3 days is probably really good.

---

It started about a week ago, BOINC would Report completed tasks to SETI but would not fetch new work.

1/24/2010 10:58:13 AM SETI@home Sending scheduler request: Requested by user.
1/24/2010 10:58:13 AM SETI@home Reporting 1 completed tasks, not requesting new tasks
1/24/2010 10:58:18 AM SETI@home Scheduler request completed

I updated BOINC to the latest Rev with out success.

I tried turning the fetch function off then on but no change.

I have now run out of work for SETI.

My preferences are set to 10 day work buffer with a connect each day.

I am still getting work from Milkyway but it is limited to only 6 work units with no regard for processing time.

Whats gone wrong?

Why won't BOINC fetch work from SETI?

It's looking as if "the latest" revision of BOINC (v6.10.29 at the time you posted) has a serious work fetch bug - introduced in the course of trying to fix more minor bugs in previous revisions.

I think I've (narrowly) persuaded Rom not to make v6.10.29 the "recommended" release on Friday (tomorrow, in my part of the world): watch out for a v6.10.31, which we should try to test quickly to make sure there aren't any errors as blatant as this one turned out to be.

Congratulations on spotting and reporting a genuine problem with only your third post on this board.

My preferences are set to 10 day work buffer with a connect each day.

I am still getting work from Milkyway but it is limited to only 6 work units with no regard for processing time.

I feel sorry for the manager trying to honour resource shares when there is a large cache & more than one project.

---

Grant
Darwin NT

My preferences are set to 10 day work buffer with a connect each day.

I am still getting work from Milkyway but it is limited to only 6 work units with no regard for processing time.

I feel sorry for the manager trying to honour resource shares when there is a large cache & more than one project.

As I understand 6.10.29, with "connect every 1 day" and "10 extra days" BOINC will try to keep 11 days cached, but only from the project with the least debt (in this case, Milkyway).

It won't resort to downloading other projects until the cache gets down to one day, or debt is "paid off."

... and if Milkyway has short deadlines (I don't crunch it) then BOINC will be reluctant to download more work if there is deadline pressure (because the deadline is effectively set when work is fetched).

In other words, it sounds like what BOINC is doing is reasonable, especially given the short Milkyway deadlines.

---