Someone else has a problem with the schdeuler going into edf mode too much? short term fix put "no dial up scheduling" to stop 2x connect rate rule.

Why does it matter if the system goes into EDF? Mine are in EDF nealy all the time.

Would like consistent use of computer. I am being forced to only have one project for 2 1/2 months because of the scheduler. If the 2x connection rate was not coming on. I would be running all three projects evenly and all three would have even usage at all times.

I do not like a scheduler that operates in its error mode constantly. I also guess you are always scratching a chalk board all the time too.

EDF is not an "error mode."

It exists because some projects have (or had) short deadlines, and round-robin scheduling won't always return work on-time if the cache (connect every 'x' days) is bigger than the shortest deadline.

If you run one project, you'll never go into EDF. If you run several projects, you can run a small cache and never go into EDF. If you run several projects and a big cache, EDF makes sure you meet deadlines, and long-term-debt takes care of dividing work according to your resource share over the long term.

But then again, ksnash, you've been told this before. You can stop worrying about EDF and everything will come out the way you want, or you can set your cache so that EDF isn't needed.

---

Someone else has a problem with the schdeuler going into edf mode too much? short term fix put "no dial up scheduling" to stop 2x connect rate rule.

Why does it matter if the system goes into EDF? Mine are in EDF nealy all the time.

Would like consistent use of computer. I am being forced to only have one project for 2 1/2 months because of the scheduler. If the 2x connection rate was not coming on. I would be running all three projects evenly and all three would have even usage at all times.

I do not like a scheduler that operates in its error mode constantly. I also guess you are always scratching a chalk board all the time too.

EDF is not an "error mode."

It exists because some projects have (or had) short deadlines, and round-robin scheduling won't always return work on-time if the cache (connect every 'x' days) is bigger than the shortest deadline.

If you run one project, you'll never go into EDF. If you run several projects, you can run a small cache and never go into EDF. If you run several projects and a big cache, EDF makes sure you meet deadlines, and long-term-debt takes care of dividing work according to your resource share over the long term.

But then again, ksnash, you've been told this before. You can stop worrying about EDF and everything will come out the way you want, or you can set your cache so that EDF isn't needed.

Is there an error that the scheduler thinks it has to fix.
Error equals too many workunits downloaded because a twice connect rule is broken.
Does scheduler enter special mode to supposedly fix this problem.
Let see. Scheduler decides that only that project that supposedly downloaded too much can run. Round robin is turned off.
Will it make you feel any better if I say it is in a constant error correcting mode.

Let's see. So people won't need extra cache they should connect to more than one project so the other project can be a backup when you can't contact setiathome. I have 3 projects. Setiathome which I can't connect to right now because I have to make up time lost by scheduler for CPDN. Einstein has had connect troubled according to messages.

---

Is there an error that the scheduler thinks it has to fix.
Error equals too many workunits downloaded because a twice connect rule is broken.
Does scheduler enter special mode to supposedly fix this problem.
Let see. Scheduler decides that only that project that supposedly downloaded too much can run. Round robin is turned off.
Will it make you feel any better if I say it is in a constant error correcting mode.
<snip>

Actually,
"There is a condition that the scheduler thinks it needs to respond to." - and it does so in the observed manner.
No error there...

---

Is there an error that the scheduler thinks it has to fix.
Error equals too many workunits downloaded because a twice connect rule is broken.
Does scheduler enter special mode to supposedly fix this problem.
Let see. Scheduler decides that only that project that supposedly downloaded too much can run. Round robin is turned off.
Will it make you feel any better if I say it is in a constant error correcting mode.
<snip>

Actually,
"There is a condition that the scheduler thinks it needs to respond to." - and it does so in the observed manner.
No error there...

There is an error it has detected

1/3/2006 6:13:01 PM||Suspending work fetch because computer is overcommitted.
1/3/2006 6:13:01 PM||Using earliest-deadline-first scheduling because computer is overcommitted.

Computer is overcommitted is not an error condition.
it is in this condition 95% of the time.
I am 100% in all cases the computer is not overcommitted and can finish even when the scheduler downloads too much. if it would not go into its edf mode.

---

Actually,
"There is a condition that the scheduler thinks it needs to respond to." - and it does so in the observed manner.
No error there...

There is an error it has detected

1/3/2006 6:13:01 PM||Suspending work fetch because computer is overcommitted.
1/3/2006 6:13:01 PM||Using earliest-deadline-first scheduling because computer is overcommitted.

Computer is overcommitted is not an error condition.
it is in this condition 95% of the time.
I am 100% in all cases the computer is not overcommitted and can finish even when the scheduler downloads too much. if it would not go into its edf mode.

I got just one suggestion, I dunno if you tried it or not yet ksnash, but have you ever thought about lowering your cache, so that you don't run into this "supposed" error your getting? Ever thing about a 1 day cache, since your run a couple different projects according to a post you posted? Each project has different deadlines. No wonder your in an EDF mode. Just a suggestion, doesn't mean you'll even concider it, since nothing works right. I'll go back to crunching my 5 projects happily with a 1 day cache now.

Jeremy

---

Is there an error that the scheduler thinks it has to fix.
Error equals too many workunits downloaded because a twice connect rule is broken.
Does scheduler enter special mode to supposedly fix this problem.
Let see. Scheduler decides that only that project that supposedly downloaded too much can run. Round robin is turned off.
Will it make you feel any better if I say it is in a constant error correcting mode.
<snip>

Actually,
"There is a condition that the scheduler thinks it needs to respond to." - and it does so in the observed manner.
No error there...

There is an error it has detected

1/3/2006 6:13:01 PM||Suspending work fetch because computer is overcommitted.
1/3/2006 6:13:01 PM||Using earliest-deadline-first scheduling because computer is overcommitted.

Computer is overcommitted is not an error condition.
it is in this condition 95% of the time.
I am 100% in all cases the computer is not overcommitted and can finish even when the scheduler downloads too much. if it would not go into its edf mode.

If you'd set the "connect every 'x'" setting lower, BOINC would not download as much.

But, let's set that aside. The scheduler has recognised that there is a possibility that work, if it takes the maximum estimated time, would not be reported on time.

So, rather than risk the possibility of a real error (by not finishing on time) it's going to do the work that needs to be done soonest first.

Yet again, the only issue is that BOINC is not processing one project for an hour, then a different project for an hour, then the next project, and back to the first -- and we're ignoring the fact that if you add up the hours at the end of the month, they'll hit the resource allocation pretty close.

... and not miss any deadlines.

---

OK,,,, I'm driving down the road.....minding my own business and the "check engine" light comes on....Is this an error or a warning light???? My car still runs, but I'd better see a mechanic soon. So I change my driving directions to the nearest mechanic instead of the original destination. It's NOT an error, but a warning condition


OR


I'm driving down the road at my usual 15 over speed limit (27 over for Kph viewers) and my fuzz buster light comes on and beeps is that an error light???? NO...It's a slow your butt down before the big bad error happens light.

Is there an error that the scheduler thinks it has to fix.
Error equals too many workunits downloaded because a twice connect rule is broken.
Does scheduler enter special mode to supposedly fix this problem.
Let see. Scheduler decides that only that project that supposedly downloaded too much can run. Round robin is turned off.
Will it make you feel any better if I say it is in a constant error correcting mode.
<snip>

Actually,
"There is a condition that the scheduler thinks it needs to respond to." - and it does so in the observed manner.
No error there...

There is an error it has detected

1/3/2006 6:13:01 PM||Suspending work fetch because computer is overcommitted.
1/3/2006 6:13:01 PM||Using earliest-deadline-first scheduling because computer is overcommitted.

Computer is overcommitted is not an error condition.
it is in this condition 95% of the time.
I am 100% in all cases the computer is not overcommitted and can finish even when the scheduler downloads too much. if it would not go into its edf mode.

If you'd set the "connect every 'x'" setting lower, BOINC would not download as much.

But, let's set that aside. The scheduler has recognised that there is a possibility that work, if it takes the maximum estimated time, would not be reported on time.

So, rather than risk the possibility of a real error (by not finishing on time) it's going to do the work that needs to be done soonest first.

Yet again, the only issue is that BOINC is not processing one project for an hour, then a different project for an hour, then the next project, and back to the first -- and we're ignoring the fact that if you add up the hours at the end of the month, they'll hit the resource allocation pretty close.

... and not miss any deadlines.

I only set it for three days. scheduler downloads upto 2 weeks worth of work. The computer is not overcommitted. It has never even been close recently of missing a deadline. Well the scheduler could cause CPDN to be late and it is the scheduler causing the problem. What the scheduler is saying is completely and utterly wrong. If it does not go in to EDF no workunit will be late. Based upon the twice connect rate.

---

WinterKnight are you still around. Have I said anything AntiBoinc. I would like something that is incorrect to be fixed. My computer is not overcommitted. The scheduler is causing a pronlem in CPDN. I have let the scheduler run without messing with it. What it says and what it does is incorrect.

Computer is overcommitted and its not.
EDF is continuous and pushes off work unnecessarily.

---

ksnash, how about just rereading the thread P@H has taken over my machine, or Can we please have control of what runs and when in which we already determined that the scheduler was doing what you set in your preferences. According to "Boincstats" you have done 56.2% CPDN, 22.8% Einstein and 20.9% seti.

Do we really need to retype the 300+ lines from the previous threads only to have you jump up and down and swear it's broke?

Posted twice for effect

ksnash, how about just rereading the thread P@H has taken over my machine, or Can we please have control of what runs and when in which we already determined that the scheduler was doing what you set in your preferences. According to "Boincstats" you have done 56.2% CPDN, 22.8% Einstein and 20.9% seti.

Do we really need to retype the 300+ lines from the previous threads only to have you jump up and down and swear it's broke?

Posted twice for effect

Everything you said is useless. You keep saying it works correctly. When what it says is wrong. It is so nice you like to support substandard software.

---

Yes, the proper use of terms is important. So let's make sure EVERYONE quits using the term "work unit" when it's really "result".

A WU is what seti splits from the Arecibo tape, the WU is then split into 4 copies and issued as a "Result" to each host. Then each result is returned to seti, validated and turn into a "Canonical result".

WE don't crunch Work Units, we crunch results.

Hows that for funky terminology???

OR

If Seti says "the uploads are failing" do they mean our downloads???

Why does General Motor put a "check engine" light indicator on the dash, by golly, they should put one light for each condition "check PFE sensor", "check oil pressure", "check air filter", "EGR valve not opening", etc etc

Yes, the proper use of terms is important. So let's make sure EVERYONE quits using the term "work unit" when it's really "result".

A WU is what seti splits from the Arecibo tape, the WU is then split into 4 copies and issued as a "Result" to each host. Then each result is returned to seti, validated and turn into a "Canonical result".

WE don't crunch Work Units, we crunch results.

Hows that for funky terminology???

OR

If Seti says "the uploads are failing" do they mean our downloads???

If they are results why are we working on them, we already have results. results are work answered.

---

They are the results of the splitters work.

If you shovel manure out of a stall, what is the result of your work? A clean stall? A manure pile in a different spot??

Why does General Motor put a "check engine" light indicator on the dash, by golly, they should put one light for each condition "check PFE sensor", "check oil pressure", "check air filter", "EGR valve not opening", etc etc

I hate to tell you, but The Ford Windstar warnings are check oil, Time to replace filter. Its funny even with my Chevy Astro you can get the people at Autozone to hook up a little device and it tells you exactly where it is failing. There has been less oxygen sensor errors recently in newer cars because the computer automatically adjust the carberator so there is no excess oxygen for sensor to activate on.

---

Yes, I'm still here, I still think, and now even more so, that the 2 * connection rule thats enforces EDF is causing problems.

This computer since the first post has now downloaded work from Einstein. One unit for crunching by Einstein 4.79 and lots of new Alberts. On the original estimate before any Alberts were crunched it looked like the system had downloaded the correct amount of work, 4.33 days worth. And then it crunched the first Albert, and it revised the estimated crunch times, I then had 140 hours worth of Einstein work on my computer, nearly 6 days.

Add that to the Seti work and I had over ten days of work, no problem I won't miss any deadlines, but the manager is still going in and out of EDF, the real silly thing every time it goes into EDF it does the Seti units, because they are due back first, and totally ignoring the Einstein units, which are the ones that have caused the problem and if there is a problem returning units on time will be the ones that will have to be aborted.

As I see it, and I see Ksnash agree's with me, If you set a connect to network setting, of even a few days, for whatever reason, The 2 * connect rule only screw's things up.

It may in the long term honour your resourse share but it does it a most perculiar way, of forcing EDF, causing -ve LTD, only doing one project, going into EDF (WHY?), hiccup at UCB, allows downloading/crunching of second project, then going in and out of EDF again on 2 * rule.

It also means that the steady flow of downloads from seti has now stopped, I haven't had any new work from Seti for nearly two days. Even though there is less than three days of Seti work in my cache, so when I do get work, one it will probably be a large batch, which can create other problems.
[i]And I'll probably be pulling down everybodies granted credits because they'll return earlier than planned for. Another current thread suggests average/unit is over 23, my average is under 20.[i]

---

Andy when did you switch to "crunch3rs'" optimized apps/cc? were you using a V4 client immediately prior to that or nearly so?

I see that somewhere between Dec 10th and Dec 13th you switched your Pentium M to crunch3rs stuff. you should have by now done enough so that the DCF would be close to accurate. This should have allowed you to download more work than you were accustomed to. I have in the past "suspended" a project that had work, and the next project then refilled my cache, and then I unsuspended my original project. I was then all set with 2x my cache. I got them done but it was close. did this sort of thing happen?

I see that somewhere between Dec 10th and Dec 13th you switched your Pentium M to crunch3rs stuff. you should have by now done enough so that the DCF would be close to accurate. This should have allowed you to download more work than you were accustomed to. I have in the past "suspended" a project that had work, and the next project then refilled my cache, and then I unsuspended my original project. I was then all set with 2x my cache. I got them done but it was close. did this sort of thing happen?

So Boinc doesn't like Pentium M's?

---