To avoid zero mark. -.-"
Besides, result become don't need when there is 3 result uploaded. This function can save more cpu cycle to do more meaningful project.
Is my opinion work.
Have a nice day.
Copyright © 2024 Einstein@Home. All rights reserved.
Could BONIC abort late result automatically?
)
Hi,
Until a deadline is actually passed, it would be dangerous for BOINC to assume that it could caclulate with certainty that a result will miss the deadline. Even if it could, there are plenty of examples of late results which actually have counted because the quorum had not been formed at the time the result was returned.
These days, the newer versions of BOINC are actually getting quite adept at managing your existing work so as not to miss deadlines and also at taking proactive action to prevent you from downloading work which may miss the deadline in the first place.
I had a quick look through your results list and there were no results visible which were in deadline trouble. You don't seem currently to have any such zero scores. I would agree with you that it is a waste if a result is going to miss a deadline but I would strongly disagree that to crunch a result that will come in after the first three have formed the quorum (but still before the deadline) is also a waste. In many instances that extra result sent out, may be needed when one of the first three fails to return for some reason. Nobody should be penalised for being part of a group of four where the first three will confirm and validate the result. Someone always has to be last.
On reviewing your results list, I have a couple of questions to ask and suggestions to make which may assist in making your contribution more efficient. I notice that you have aborted a number of results (including one with a small amount of elapsed cpu time) for some reason, but not because of deadlines according to the dates displayed. Can I ask why you chose to abort those? Also can I ask if you are supporting other projects?
Currently, the website has 8 results listed as being "In Progress" but I strongly suspect that you can only see two under your work tab in BOINC Manager. Can you please confirm this? If this is true, I can understand that you may have seen this type of result eventually expire in the past. You were probably concerned at the apparent waste and thought that something needed to be done about it. Can you tell us if this has been happening previously?
Once you give us the answers to these questions and any other information you feel is relevant, I'm sure we can come up with a plan to stop this happening in the future.
Cheers,
Gary.
Dear Gary, Thank you so
)
Dear Gary,
Thank you so much for your reply. Totally understand and agree you explained and your opinion.
As you saw in the web, I haven't got the deadline problem, just a brain storming by seeing other WU got this condition. So try to ask could BONIC Manager discover and delete the result that after deadline.
For the questions you asked, I am the member of P@H. For the aborted results, I think it is related to my stupid computer, it will freeze at any time, any condition just according its mood. >__< I've try to reinstall my OS but it still appear.-.-" Is my Chinses Window XP conflict with BONIC?
For the question about the different number of result appear between web and in work tab in BOINC Manager, I've got the rrght number this time. But in fact, I've to meet the condition you explained, but just don't know how to ask. :P
Are you know the reason?
Thanks again for your kindly reply.
Finally, sorry for my poor English if there is any sentenses you can't see clearly. ^^"
Have a nice day
Louis
RE: As you saw in the web,
)
I notice that you have aborted a result recently (ResultID=10101763) where it was about 85% done and it was going to be the last one returned as the quorum had already been formed. If you are really so determined to abort your results if they wont be one of the first three to be returned, your better option would be to reduce your "connect to network" interval so that you only keep "fresh" work on your computer. If you do that, you can virtually guarantee that you will always be within the first three to finish.
I don't understand what you mean by "force me to abort". Every 60 seconds while a result is crunching, a checkpoint file is written so that crunching can restart from a known point in the event of a windows crash. Normally, when a machine is rebooted, BOINC is able to restart the science app which will pick up its start point from the checkpoint file. There should be nothing "forcing" you to abort. If the computer is unstable because of hardware issues such as faulty RAM, overclocked cpu, overheating problems, etc, then I guess data corruption is likely to occur, causing client errors and loss of results. However, when I look at your client errors, most of them seem to be you making a conscious decision to abort - ie I see "Aborted via GUI" as the message.
I wouldn't have thought that there would be a problem between BOINC and the OS. I see you only have SP1 for your Windows XP. Perhaps you could install SP2 and see if that gives you a more stable system. Your crashes sound more hardware related though. You could try running something like Prime95 in torture test mode. That would soon see if your hardware was stable.
Yes, the reason for ghost work units is that the server sends data from time to time which the client (for largely unknown reasons) fails to receive. The server thinks the client has received the data and so it appears in the list of results on the website. The client however doesn't actually have the work. This has now been fixed by implementing a "handshake" protocol so that server and client check with each other from time to time to keep their lists of work in sync with each other. If you ever have a problem with significant ghost work units, you should probably upgrade to 5.2.2 to get this (and other) benefits.
At the moment, you only have three results in progress and if you can see three in BOINC Manager then you don't have a ghosts problem. The other day when you had 8 results in progress, I thought you may have had a ghosts problem.
So, can you please let us know what value you are using for "connect to network" setting?
Cheers,
Gary.