You're right but projects managers prefer probably resent automatically a few WUs instead of managing deadline extension or WUs cancellation.

Please tell me how the project managers would have to deal with "managing deadline extension or WUs cancellation" any more than they would have to deal with managing sending work to client computers.

WOULDN'T THIS BE AUTOMATED? Or did you seriously think that I expect someone to sit in a room and pine over how to deal with each circumstance..?

Whooptie-doo, 2, maybe 3 extra settings. I.E....

1. Should the client be allowed to extend work, or should expired work be aborted?

2. If EXTEND at above, please indicate percentage of original deadline to extend work.

3. How many times can a client request for an extension?

(IMHO any work that has coninical verified results already should be aborted.)

OMG that's just soo many questions for a project manager to deal with.

---

Because at the time a WU is overdue it is automatically resent to another user. I let you explain how to manage deadline extension.

---

Because at the time a WU is overdue it is automatically resent to another user. I let you explain how to manage deadline extension.

Obviously, this doesn't happen immediately.

If the work has been already resent to another host, then abort it. big deal.

---

"Obviously, this doesn't happen immediately."

That's not an answer.

---

Well, if the WU is past the deadline, would not a new sent of a ResultsID delay the result or WU even more, if the previous ResultID is only delayed (but still alive).
I mean - the problem with long WUs are their 'real' ongoing status. Perhaps 'Enhanced' SETI should have taken some knowledge from CPDN and introduce trickle-up message (sending partial result or you just a message to a server "I'm alive and at xx% of the damn WU"). One packet do the job...

I wonder how many WUs there will be 'in progress' for weeks or months - until server gets the real status (report, error, past deadline etc.) when Enhanced is introduced in large scale.

There is also a possibility to terminate WU from a server side (we call in trickle-down message on CPDN) but I would be carefull with such feature. [It still make sense on CPDN, where you want to terminate model thay is way off or you are no longer interested in this particular model due to it's parameters or whatever].

I aggre with Aaron that BOINC was not ment to be babysitted.
On the other scale, there is another extreme - one may manually trasfer a WU that looks like to go beyond deadline on another machine. It is not a trivial step but it can be done (suppose some knowledge of BOINC and it's xml structure). But again - BOINC is suppose to deal with deadline, buffer etc.

---

Jesus christ people, use your damn heads. {snip} It was designed as a screen-saver. The type of program that you install and FORGET ABOUT. Have we lost track of this?

Screen-saver program? Are you sure?

Classic may have been, but BOINC never was. It's an option. Hence why some projects don't use it. Might be me, but I research things before I call out Christ's name and then blame my misreadings on others. And I am not even that religious!

BOINC is getting to be fire and forget. If you just wait a little more until the AMS (Account Management System) is fully integrated. It is being tested at this moment, what more do you want? Test it yourself? Go make one then!

It can also be said, that GENERALLY, work would probably be useless the farther you get past the deadline.

Outside the deadline, yes. But inside the deadline, what makes you say so? Because you are the 4th, 5th .. 9th to return the result you crunched on? It's still saved to the database. It's even saved to the database when you are outside the deadline, but when the validator is offline, just so you can get your data back!

With these projects, if you have a computer that FOR ONE REASON OR ANOTHER gets shut off for a time, your computer will CONTINUE TO CRUNCH PAST THE DEADLINE. This is a serious waste of processing power.

No, once your computer for whatever reason didn't return the data within the deadline, the unit gets send to another computer which hopefully can send its data back within the deadline.

Redundancy is there. Just look up any of your crunched units at Seti, and see at the top what the redundancy is. like this one of mine: minimum quorum = 3; initial replication =4; max # of error/total/success results 5,10 ,5

When the computer comes back online, it will continue to crunch the EXPIRED AND ALMOST ALWAYS USELESS work, until it is gone. The only way to overcome the above scenarios without crunching for DAYS doing useless work, is to either..

1. Manually abort the work.
2. Include some sort of method for the core to deal with work past (or swiftly approching) it's deadline.

The methods that I propose provide for a user-friendly, and intervention free method of querying the server on wether or not the work should be ABORTED or the deadlines should be EXTENDED.

Why not have BOINC look at the deadlines and look at the actual time/date that the computer lives in and have it delete the unit(s) that is/are really past its/their deadline?

Do you ask your supermarket to extend the "best by" date on your milk, just because you like to have it in your fridge? Or do you throw it out?

---

With these projects, if you have a computer that FOR ONE REASON OR ANOTHER gets shut off for a time, your computer will CONTINUE TO CRUNCH PAST THE DEADLINE. This is a serious waste of processing power.

No, once your computer for whatever reason didn't return the data within the deadline, the unit gets send to another computer which hopefully can send its data back within the deadline.
Redundancy is there. Just look up any of your crunched units at Seti, and see at the top what the redundancy is. like this one of mine: minimum quorum = 3; initial replication =4; max # of error/total/success results 5,10 ,5

How in the hell does your answer to my statement equal NO? If I say that the computer that gets shut off will crunch a unit past it's deadline and this is a waste of cycles, and you say "NO - The unit gets sent to another computer." How in the hell does sending out YET ANOTHER workunit not amount to a waste of CPU cycles on the first computer when it processes the work past it's deadline when it's not needed?

Even further.. If the server is sending out the workunit AGAIN to cover for losing your computer's allocated workunit, and the quorum is 3, then having 4 results is redundant and wasteful - OTHERWISE THE QUORUM WOULDN'T BE 3. However - Not every project has a quorum of 3. Some need only 1 result per W/U. (like Rosetta) IN EITHER CASE, processing work past it's deadline when it is no longer required is wasteful. Don't skirt the issue.

Why not have BOINC look at the deadlines and look at the actual time/date that the computer lives in and have it delete the unit(s) that is/are really past its/their deadline?

Do you ask your supermarket to extend the "best by" date on your milk, just because you like to have it in your fridge? Or do you throw it out?

This is also an option, but rather than just assuming that the work isn't needed (and dumping it), I feel that asking the project server if more time is appropriate would allow for projects to keep bandwidth to a minimum as workunits then do not need to be resent. This is especially true for projects with large workunits and/or bandwidth constraints.

---

Adding to my above post:

1. Power Outage

Nothing you can do about it, but that's why most projects send out WUs in redundancy.

2. Person on vacation doesn't want the computer on while they are gone

Then you'd set your cache to minimum (0.00000001) days of contact. You'll lose at maximum one unit (per computer you shut down).

3. A part of the computer failed, and is being sent back for RMA - I.E. Motherboard RMA, or Memory RMA

Doesn't matter that much, as when it is sent back, you will most probably get a new hostID anyway.

4. Someone moves, and the computer is packed away for a while

See 2.

5. It's a laptop that someone took with them somewhere that they can't leave it running BOINC all day.

???? Stolen laptop? Given away with the program on it? A laptop can't run BOINC 24/7 at this moment anyway as it will heat up too much.

6. It's a laptop that they have shut off or suspended without thinking about it. I.E. Shutting the case on some models can result in this!

You're reaching now. See 6.

7. Different kinds of Network outages PICK ONE.

Same as 1. Out of your reach, unless you broke the network and have to fix it. If corporate network, they will usually be up within an hour or they lose too much money.

8. Network administrator restarts the PC via the LAN, but neglects to log the user in, on a single user style installation.

Too bad. See 1.

9. Multi-user PC's, Single-user installations in which the user that installed BOINC does not use the machine for days.

For 8. and 9. the answer is simple: Service install where possible. Else see 1.

---

In answer to every single answer in the message below.. (message 1591)

THIS IS NOT HOW BOINC WAS ENVISIONED. AT ALL.

The attitude you have of "too bad, sucks to be you", "the problem doesn't exist" and "micromanage everything you do OR DON'T PARTICIPATE" is NOT why BOINC was created.

I fear for any BETA project you contribute to.

---

With these projects, if you have a computer that FOR ONE REASON OR ANOTHER gets shut off for a time, your computer will CONTINUE TO CRUNCH PAST THE DEADLINE. This is a serious waste of processing power.

No, once your computer for whatever reason didn't return the data within the deadline, the unit gets send to another computer which hopefully can send its data back within the deadline.
Redundancy is there. Just look up any of your crunched units at Seti, and see at the top what the redundancy is. like this one of mine: minimum quorum = 3; initial replication =4; max # of error/total/success results 5,10 ,5

How in the hell does your answer to my statement equal NO?

Look mister... I have had it with your language. Do you want a discussion, or do you want to have a fight? Do you want a solution to what you see as a problem, or do you only want your solution to where there is no problem?

Do you read things or only skim them and find fighting points? For if you do, I am out of the discussion and you're on my ignore list. Simple as that. Read my post below/above this one. It explains things.

If I say that {snip} it's not needed?

You didn't read my whole post. You just went ahead and posted.

Why not have BOINC look at the deadlines and look at the actual time/date that the computer lives in and have it delete the unit(s) that is/are really past its/their deadline?

Bottom of this post. That was what I was giving as an option. Please read before you furiously attack.

Even further.. If the server is sending out the workunit AGAIN to cover for losing your computer's allocated workunit, and the quorum is 3, then having 4 results is redundant and wasteful - OTHERWISE THE QUORUM WOULDN'T BE 3.

Ah, but what if the server sends out the same result to 3 computers and only 2 send a result back? Do you want to answer the people who may have to wait 2, 4, 6 weeks for their credits to be granted? Or do you want a 4th result being sent to a 4th computer, no matter if the other 3 return the result without a problem?

However - Not every project has a quorum of 3. Some need only 1 result per W/U. (like Rosetta) IN EITHER CASE, processing work past it's deadline when it is no longer required is wasteful. Don't skirt the issue.

I am not, you are. You want an extended deadline for all projects when the BOINC program on your computer asks for it. You don't like the deadline given, fine. But why should the program on your computer ever ask for an extended deadline?

As for bandwidth restrictions, again, why do you think it's sent out with redundancy? And how much bandwidth will be taken extra of all those computers "requesting" more crunch time for the units they have on?

I don't assume you have one unit per project. So how does BOINC ask for extended deadline on all of the units for all of the projects on your computer? Or is it just one unit? Or are there multiple ones, cross project?

---

THIS IS NOT HOW BOINC WAS ENVISIONED.

Oh, please... how was BOINC envisioned then, oh greatness?

Were you the birth father of the whole concept, or did I miss something? But go ahead, tell how it is supposed to be and then explain where http://boinc.berkeley.edu/ went wrong. Please. :)

---

::post deleted::

The quality of beta testers has seriously taken a nosedive in prior months.

I abscond myself from this debate for fear of needlessly infuriating myself over someone who has no clue.

---

You do know that you've been posting this on the wrong forums all this time? That these are the Seti Beta forums, which handle the transition to new Seti clients?

Now, since I don't have a clue, or so you say, I am probably completely wrong... but BOINC does all the transfering and checking of things. Not the present Seti client, nor the Enhanced Seti client.

You do know that these are the BOINC development forums? A place you would have better posted it at?

But no, I don't have a clue. :)

Thanks for the laugh. See you in Seti Q&A, where there's more of us who don't have a clue helping out those who also don't have a clue. :)

---

Why should SETI Enhanced BETA not be a focal point for this discussion, considering that the time to complete a workunit just shot up by 1000% or more?

And I am more than familiar with the BOINC Dev boards. This is also an issue that has been taken up in other areas.

---

Why should SETI Enhanced BETA not be a focal point for this discussion, considering that the time to complete a workunit just shot up by 1000% or more?

Well, since I am clueless... maybe because the deadline on Seti proper will also go up once the new client is released? It's just a hunch. Or maybe because Seti wants everyone to use the 5.2.x client, which will guestimate the end of the result more correctly if you give it time on fire and forget.

Yet you were talking about all BOINC projects, talking about BOINC. Not just the project here.

It's not the BETA test project I'm referring to, I'm talking about BOINC-wide.

your post.

What Seti is trying to do is relieve their front end and back end: Less downloads a day, while getting more data per result. Do I know for sure if the deadline will go up? No, but I suspect it will.

If I knew, would I still be clueless though?
Do you know? If not, what does that make you? :)

---

Aaron
You seem to forget about the overhead this would cause. What you are asking for would mean almost constant contact between the project server and your/my client. This contact would be so the server knows the condition of the work on your/my system. Other wise the server would not know that your system has the result almost finished and then from that info the project server would have to tell the scheduler not to create or create depending on circumstances a replacement wu for sending. And what about those on dial-up, how are their systems going to stay in almost continuous contact with the projects? Or if your/my system is off or broken or down for file maintenance there wil be no communication then either.

And another thing, I never need to micro-manage Boinc. I set it but don't forget it because I am interested. But because of all my observations of how it(boinc) manages itself I know that it will run without any need for intervention. I have not changed any settings for months and have yet to even come close to missing a deadline!

---

98SE XP2500+ @ 2.1 GHz Boinc v5.8.8

Aaron
You seem to forget about the overhead this would cause. What you are asking for would mean almost constant contact between the project server and your/my client. This contact would be so the server knows the condition of the work on your/my system. Other wise the server would not know that your system has the result almost finished and then from that info the project server would have to tell the scheduler not to create or create depending on circumstances a replacement wu for sending. And what about those on dial-up, how are their systems going to stay in almost continuous contact with the projects? Or if your/my system is off or broken or down for file maintenance there wil be no communication then either.

I hope I'm understanding this correctly, bear with me :)

- In the first instance - Where work is nearing the deadline, but the computer could not complete it before the deadline expires, I'm sorry, but I fail to see how notifying the server that additional time would be needed would create MORE overhead than making the server create more work (which then has to be sent to another client) then accepting your workunit anyway, and validating the extra work from both your computer and the (now) extra computer.

- In the second - where work is passed it's deadline, the computer still has to upload the result when it completes it. The communication would be much smaller if the work (and result) was no longer required. If another workunit was created already, but not assigned or sent out yet.. Then it could be removed, and that result would no longer be required. Wouldn't that be less overhead also?

So.. in defense of Aaron.. I fail to see your point, but again, I just got here! Also.. I don't think that this is something that could be implemented on a project level, but every project should be concerned about it. Certainly though Aaron has some valid points, because sometimes I leave my computer on for weeks, and sometimes I'll shut it off for days. I certainly wouldn't want my computer crunching worthlessly when it comes back on, and it's not something I would normally think about either. However, now that I've read this, I'll check Boinc whenever I turn the computer back on and abort whatevr work is too late. I'd rather boinc do this for me though. I really didn't know that Boinc would crunch past a deadline?

---

Ok, the overhead comes from the almost constant contact needed to do this and it would have to be for every project that you/I are connected to. THey would rather reduce the communication traffic than increase it.

---

98SE XP2500+ @ 2.1 GHz Boinc v5.8.8

But in every instance I just can't see how it would actually INCREASE traffic, only reduce.

Creating more work, and sending it out to another client equals less traffic?

explain.

---

If you plan to shutdown for what ever reason you can use the 'No New Work' button to stop getting more work. That was something that was added a few Boinc client versions back so that we could stop getting work when we want to go on holidays etc.

---

98SE XP2500+ @ 2.1 GHz Boinc v5.8.8