I don't see what's wrong with just letting it run? I've yet to have anything turned in late shy of hardware failure or user error. So maybe I'm missing something here but what exactly does it matter? Although I agree it seems most reasonable for FIFO on my machine it may not make much sense on others under certain circumstances though.

It matters in situations when unexpected things happen:
- You make hardware changes and computer stays in peaces on the table several days
- Your ISP is unreliable and you may be sometimes without Internet
- Urgently you have to go out of town and don't want to leave the computer ON
- SETI/project servers fail and do not work for a week
- ...

If you use "Computer is connected to the Internet about every 0 days" you/BOINC may miss some deadlines.

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Miep - thanks.

I had blanked out in my (alleged) mind that bit about "connect every n days" and have it set to 0. Now I know what you were talking about.

---

I don't see what's wrong with just letting it run? I've yet to have anything turned in late shy of hardware failure or user error. So maybe I'm missing something here but what exactly does it matter? Although I agree it seems most reasonable for FIFO on my machine it may not make much sense on others under certain circumstances though.

It matters in situations when unexpected things happen:
- You make hardware changes and computer stays in peaces on the table several days
- Your ISP is unreliable and you may be sometimes without Internet
- Urgently you have to go out of town and don't want to leave the computer ON
- SETI/project servers fail and do not work for a week
- ...

If you use "Computer is connected to the Internet about every 0 days" you/BOINC may miss some deadlines.

In every one of your situations, BOINC has recovery mechanisms and metrics in place (DCF, Time powered on vs. time allowed to crunch, amount of time internet service is available).

BOINC will see that time has passed since you last powered on (in the event of a power-down; in pieces all over a table, or powered down because you left town), and will adjust itself accordingly. More than likely it will start running in EDF mode if enough time has elapsed and the simulations show it is required.


Even disregarding all of that. I've had servers go down for over a week and I don't micromanage their caches and workloads. If it finishes a workunit on time, then it finishes it ontime and the built-in safeguards work. If it doesn't, it's not a real loss. The workunit will be sent out to someone else and crunched and the work will get done. It's not a big deal.

If people really want to micromanage BOINC because they think they can do it better, fine. No one can really stop them. But I don't think BOINC needs to be redesigned because a minority of users have enough spare time on their hands to manually do what the computer does just fine if left alone.

In every one of your situations, BOINC has recovery mechanisms and metrics in place (DCF, Time powered on vs. time allowed to crunch, amount of time internet service is available).
BOINC will see that time has passed since you last powered on (in the event of a power-down; in pieces all over a table, or powered down because you left town), and will adjust itself accordingly.

Only if these unexpected things happened regularly in the past,
but in this case they are not unexpected ;)

If people really want to micromanage BOINC because they think they can do it better, fine.

Setting "Computer is connected to the Internet about every 5 days"
and "Maintain enough work for an additional 0 days"
(instead of 0+5) is not micromanaging - it is just config.

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

In every one of your situations, BOINC has recovery mechanisms and metrics in place (DCF, Time powered on vs. time allowed to crunch, amount of time internet service is available).
BOINC will see that time has passed since you last powered on (in the event of a power-down; in pieces all over a table, or powered down because you left town), and will adjust itself accordingly.

Only if these unexpected things happened regularly in the past,
but in this case they are not unexpected ;)

Try it. Randomly take your machine offline for a week or even two weeks, then put it back online again. Watch how gracefully BOINC is able to recover.

I had two personal servers go offline: one for over a month and one for a week. I had to abort the overdue tasks on the one that was offline for over a month, but the one that was offline for a week was able to recover without issues.

No other manual intervention is required.

If people really want to micromanage BOINC because they think they can do it better, fine.

Setting "Computer is connected to the Internet about every 5 days"
and "Maintain enough work for an additional 0 days"
(instead of 0+5) is not micromanaging - it is just config.

Sure, it is an incorrectly set configuration if the computer connects to the internet more often than every 5 days.

But I fail to see how it isn't micromanaging if one attempts to manually intervene in the work order process or the work fetch process. BOINC can still recover gracefully even in the event of a misconfiguration.

Try it. Randomly take your machine offline for a week or even two weeks, then put it back online again. Watch how gracefully BOINC is able to recover.

No need to try - I know it. BOINC learns and adapts.
But no-one can adapt to unexpected. If you have earthquakes every day - you adapt, if no earthquake for 100 years - it is unexpected event end nobody is prepared.

If BOINC knows from the past "experience" that the computer is ON all the time and Internet works all the time
and you have tasks with deadline tomorrow if you "take your machine offline for a week or even two weeks" BOINC will miss the deadlines.
And some may even have 100s of tasks completed but not reported when the "unexpected" happens.

I don't like if my computers' work go to waste, you may be different.

Sure, it is an incorrectly set configuration if the computer connects to the Internet more often than every 5 days.

Why incorrectly??
BOINC connects whenever it likes but is prepared for the unknown bad future (of 5 days ;) ).

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Try it. Randomly take your machine offline for a week or even two weeks, then put it back online again. Watch how gracefully BOINC is able to recover.

No need to try - I know it. BOINC learns and adapts.
But no-one can adapt to unexpected. If you have earthquakes every day - you adapt, if no earthquake for 100 years - it is unexpected event end nobody is prepared.

If BOINC knows from the past "experience" that the computer is ON all the time and Internet works all the time
and you have tasks with deadline tomorrow if you "take your machine offline for a week or even two weeks" BOINC will miss the deadlines.

You may want to check your schedules dude. Mine is set to connect every .25 days with a 10+10 cache. I keep around 2500-3000 work units on hand at a time, I often turn my main cruncher off for a few days to a week at a time during my busy times and never have late turn ins. When they send you an AP unit you get a 25 day schedule, MB's are 14 days, least on my machines. In any given 24 hour time span my machines clear through a good bit of work. I could turn both of them off for a week and it wouldn't matter. Now stretch out to 1.5 weeks and I would start having things run late.

Unexpected things will happen, however that's why you have measures in place to take care of it. It's why we have resends, wingmen, and schedulers. It just doesn't matter, just let it do it's thing and quite worrying about it. If your machine isn't getting things done on time is a PIBKAM problem. heh.

---

Traveling through space at ~67,000mph!

Mine is set to connect every .25 days with a 10+10 cache.

Now that is impossible - .25 days & 10+10 cache contradict!

If your machine isn't getting things done on time ...

Where did you get such impression?
My machines return work 20-30 days before the deadline (10+10 cache).

(If you have 10+10 cache then you do what I propose - "connect every 10 days" makes your BOINC finish the task and report it 10 days before the deadline)

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Try it. Randomly take your machine offline for a week or even two weeks, then put it back online again. Watch how gracefully BOINC is able to recover.

No need to try - I know it. BOINC learns and adapts.
But no-one can adapt to unexpected. If you have earthquakes every day - you adapt, if no earthquake for 100 years - it is unexpected event end nobody is prepared.

If BOINC knows from the past "experience" that the computer is ON all the time and Internet works all the time
and you have tasks with deadline tomorrow if you "take your machine offline for a week or even two weeks" BOINC will miss the deadlines.
And some may even have 100s of tasks completed but not reported when the "unexpected" happens.

BOINC has a few stats that cover some unexpected occurances. "Percent of time BOINC is running", "While BOINC running, percent of time work is allowed", and "While BOINC running, percent of time host has an Internet connection" are all immediately affected, even during a one-time, random event. BOINC will noticed that wall time has passed since it last crunched and it will turn over to EDF mode to try to bail itself out of the situation, and it will stop requesting new work until it feels "safe" again.

Sure, there's a possibility of missing some deadlines, but even manual involvement is not going to do any better if there simply isn't enough time.

I don't like if my computers' work go to waste, you may be different.

If you've missed too much time, no amount of manual involvement is going to avoid waste short of aborting anything that is due immediately and letting your computer work on what's left.

If you're really worried about "waste" then you should make sure your computers have a purpose being on in the first place, and that you're donating spare cycles, not extra, created cycles. My servers would be on anyway, so they may as well crunch.

Sure, it is an incorrectly set configuration if the computer connects to the Internet more often than every 5 days.

Why incorrectly??
BOINC connects whenever it likes but is prepared for the unknown bad future (of 5 days ;) ).

Why incorrectly? Because you shouldn't be setting that your computer won't be connected to the internet but every 5 days if it's connected all the time. If you're worried about your computer crashing, then set the cache for 5 days, not the "Computer is connected..." setting. The "Computer is connected..." setting is for dial-up users that aren't permanently connected to the internet and may be 5 days without access. "Computer is connected..." was never intended for "unknown emergency situations" such as power failure or computer crashes. If that's what you're using it for, then yes, you are using it incorrectly.

Mine is set to connect every .25 days with a 10+10 cache.

Now that is impossible - .25 days & 10+10 cache contradict!

If your machine isn't getting things done on time ...

Where did you get such impression?
My machines return work 20-30 days before the deadline (10+10 cache).

(If you have 10+10 cache then you do what I propose - "connect every 10 days" makes your BOINC finish the task and report it 10 days before the deadline)

Well guess what that's the way it's been set, since the beginning. I always keep about 1800-3500 tasks between machines and never run late. It phones home 4 times a day to request work/ask for work etc. to keep a 10 cache running plus an additional 10 days. .25 "/" 10+10 is both machines. LEft out the bolded part. As far as when my work is getting done my machine is currently working on tasks due 4/20, 4/26, 5/21-22-23-24. So no not every 10 days got some not being worked on now that are due in 7 days. Each to his own the scheduler hands you tasks with 14 and 25 day due dates, if your machine turns them in late it's not due to BOINC. As far as your 10 day rule you should watch your computer more, mine uploads most every time it gets done with a task, and asks for work just about as quickly. Oh and I never said you were late, quit trying to make it personal sheesh.

---

Traveling through space at ~67,000mph!

Should you be setting that your computer won't be connected to the Internet but every 5 days if it's connected all the time?

Yes

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Should you be setting that your computer won't be connected to the Internet but every 5 days if it's connected all the time?

Yes

That's a matter of opinion. No substance.

---

Traveling through space at ~67,000mph!

Should you be setting that your computer won't be connected to the Internet but every 5 days if it's connected all the time?

Yes

The correct answer would be "no". You should not be setting "Connect to..." to anything more than 1 if you are permanently connected to the internet.

Should you be setting that your computer won't be connected to the Internet but every 5 days if it's connected all the time?

Yes

That's a matter of opinion. No substance.

It's not even a matter of opinion. I can respect if people choose to use it for extra cache (even if I disagree with the idea), but I cannot respect if people who use it for extra cache then blame BOINC's scheduler as being poorly programmed or unnecessarily complex when they're using the setting incorrectly in the first place. If they are unaware of how the setting is used, I can forgive that. Foreknowledge is forewarned.

It phones home 4 times a day to request work/ask for work etc. to keep a 10 cache running plus an additional 10 days.

Yes, mine too. So where is the problem somebody see?
Nothing wrong happens if you/I set "Computer is connected to the Internet about every 10 days"

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

It's not even a matter of opinion. I can respect if people choose to use it for extra cache (even if I disagree with the idea), but I cannot respect if people who use it for extra cache then blame BOINC's scheduler as being poorly programmed or unnecessarily complex when they're using the setting incorrectly in the first place. If they are unaware of how the setting is used, I can forgive that. Foreknowledge is forewarned.

I agree, however I use 10/10 for the extra cache and still don't have an issue with getting work turned in on time. It also come in handy when the Seti servers decide to die for a week or so and I can continue crunching. I don't think there is a "wrong" setting. What happens is people setup the wrong settings with how often they leave their computer crunching. I mean if you setup a 5 day cache but only let your computer crunch an hour a day you will end up missing a bunch of deadlines. There is a lot more involved in turning work in on time than simply setting up Boinc's cache. You have to let it run! I do have an issue with people saying that it's Boinc's fault things get turned in late, when it simply isn't, goes back to the PIBKAM issues.

It phones home 4 times a day to request work/ask for work etc. to keep a 10 cache running plus an additional 10 days.

Yes, mine too. So where is the problem somebody see?
Nothing wrong happens if you/I set "Computer is connected to the Internet about every 10 days"

I think you and me are saying the same thing in two different ways lol. On average mine connects 4 times a day but more often than not it connects immediately upon finishing a workunit and asks for more work. So sometimes it's connecting every few minutes. Just depends on what is being worked on and how often it's got another one being completed behind it. I've got 10 units at a time going and more often than not one is only 5 minutes behind the other give or take a few minutes.

---

Traveling through space at ~67,000mph!

Should you be setting that your computer won't be connected to the Internet but every 5 days if it's connected all the time?

Yes

The correct answer would be "no". You should not be setting "Connect to..." to anything more than 1 if you are permanently connected to the internet.

Why?

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

I mean if you setup a 5 day cache but only let your computer crunch an hour a day you will end up missing a bunch of deadlines.

Not true - if you ALWAYS "let your computer crunch an hour a day" BOINC will NOT miss any deadlines, it will just get less tasks (24 times less tasks).

The setting means "5 calendar days" - NOT "5 computing days".

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

The setting means "5 calendar days" - NOT "5 computing days".

Agreed, however if your computed has been on crunching for 30 straight days, then you cut it back immediately to 1 hour a day, you will miss some deadlines until the scheduler and your computer recompute what you can get done. In that effect it's not Boinc's issue it's yours.

---

Traveling through space at ~67,000mph!

All of this arguing, yet Boinc did miss a couple of deadlines for me on Einstein last week, My systems are always on and I did not do any manual intervention, I wanted to see what would happen. The tasks were downloaded to my system on March 23, worked to about 50-ish percent, had an accurate estimate of the time left on them, then suspended until a couple of hours before deadline, on April 6. They resumed with less time left to deadline than Boinc was even estimating they had left to process, so what's up with that? Oh, and I use a .2 day connect interval.

-Dave

---