Message boards : Number crunching : Running out of time for a packet
Message board moderation
Author | Message |
---|---|
Send message Joined: 4 Jan 07 Posts: 9 Credit: 575,832 RAC: 0 |
I have a work unit processing which I think is going to time out before finishing. For a long time I thought it would complete within the deadline; I now can see it won't. http://climateapps2.oerc.ox.ac.uk/cpdnboinc/workunit.php?wuid=8407572 If it runs out of time: - will someone else have to process the same packet from scratch again? - is the 6 million CPU seconds wasted? - do I still get the WU credits? As far as I can make out, it has got further with this packet than the 4 wingmen did; is there any way to allow it to continue beyond the deadline? Is there a benefit in doing so? |
Send message Joined: 31 Aug 04 Posts: 391 Credit: 219,896,461 RAC: 649 |
On CPDN the deadlines are not strictly enforced. Someone more expert than I has already posted about this in other threads. Keep on crunching |
Send message Joined: 29 Sep 04 Posts: 2363 Credit: 14,611,758 RAC: 0 |
Eirik's right, and you will get the proper credit for the whole model as well. If you can, it's still a good idea to finish the model as soon as possible after the deadline because all the current models (of all types) are time segments of long time periods. When your model finishes, another computer will be able to use your data to begin the next segment. Cpdn news |
Send message Joined: 4 Jan 07 Posts: 9 Credit: 575,832 RAC: 0 |
Thank you; that's a relief. When processing packets this size it feels to me I have taken on a voluntary responsibility and I have an obligation to see it through. ;-) I just hadn't expected the PC in question to plod along as slowly as it has. (The PC in question is my 2nd desktop PC. I have been wanting to replace the operating system from Ubuntu to Linux Mint for about 6 weeks now. The only thing stopping me is waiting for this data packet to finish! I did create a Mint partition and try to move the data packet but I couldn't work out how to get Boinc to recognise it. I did this once before in Windows but cannot work it out under Linux. Never mind, I've decided I'll wait.) Edit: Having re-read the Recommended System Requirements, I see why I have this feeling of commitment (it asks for it) and that the PC is just above minimum spec. Watching this packet is like owning a Tamagotchi :-) |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,011,472 RAC: 21,368 |
I am running an even slower computer - a 1.6GHz dual core atom so I understand your dilemma! |
Send message Joined: 29 Sep 04 Posts: 2363 Credit: 14,611,758 RAC: 0 |
I ran a 160-year BBC model on a 1.33GHz machine which was below minimum spec. It completed nearly a year later. If you have a slow machine with only one or two cores I think you should take regular backups of the complete contents of the BOINC Data folder. Because the longer a task spends running on a computer the higher the likelihood that something will happen to crash it. Cpdn news |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
Atom computers are designed for low power, so as to be small and light weight. They aren't meant for the sort of high powered crunching being done here. They take every opportunity to slow down the processor to save power. While this project still accepts data long after the artificial deadline, that doesn't mean that the final results will get used. These days, the results are going directly to the research group that is involved. That's good for the Oxford people, because it gets around the need to constantly find more storage space. But these people need their data ASAP. Or sooner. So, if a computer takes way beyond the "deadline" to finish, it will still "time out" at that deadline. At which point, the server can easily re-issue that data set. And if this then gets run quickly on a fast processor, it will most likely get finished and returned before the plodder. So it will be THAT result that gets used to create the next data set in the sequence. Just a thought. |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,011,472 RAC: 21,368 |
As the other tasks in the cm3n work unit that I have one of running on the atom all failed I will let it continue - almost at 75% point. It will now only download the regional models. |
Send message Joined: 29 Sep 04 Posts: 2363 Credit: 14,611,758 RAC: 0 |
Could I suggest that for slower computers and for computers that are turned off a lot of the time it is better to crunch the shorter CPDN models and not the longer ones. You can go to your account (link in the blue menu on the left) and then in your climateprediction.net preferences you can edit the types of model you want. Hadcm3n is not a good idea. It's long. Deselect it. All the regional models are shorter. So select Hadam3p EU, SAF, PNW or ANZ. The changes will take effect after your computer contacts the CPDN server. Cpdn news |
Send message Joined: 15 May 09 Posts: 4540 Credit: 19,011,472 RAC: 21,368 |
It was user error that led to me running hadam3cns on the atom. I won't be doing any more of them on it. - it still has 837 hours left at 72% finished! I thought I had set the computer to work but hadn't so it assumed it was home and is now suffering the consequences. |
©2024 cpdn.org