Long runtimes on projects?

[darwincollins]

Application: cmsearch 0.14
Name:
cms_GA-p[ST-Lin64f]_Streptococcus-pneumoniae-70585_CP000918.lin.EMBL_RF00030_RNase_MRP_1283507736_2021_0

It says elapsed time: 158:58:40 Progress 6.686%
To completion: 415:53:52:04 Report deadline: 9/13/2010
Running, high priority
They are using less than 230MB ram. (8gb on machine)

This app is using only one 3ghz core. (whereas others like Aqua will use all 8 cores)

I have one other RNA app that also has a long way to finish.

Since I doubt it would finish by Sept 13... What should I do? Just leave it alone? That is acceptable?

While watching the completion timer count down tens of seconds within a second, I know it won't take 415hrs to complete, but should I get a unit that would take so long to complete with such a short report deadline?

Thanks

[yoyo]

Hello,

something seems to be strange with your system. This workunit was already finished by 8 hosts and the consumed cpu time is between 27 and 50h.
Keep in mind, that the application has no checkpoints. So if you shutdown your Boinc it restarts the workunit from the beginning.
You should consider to abort the wu.

yoyo

[Torbjörn Klatt]

First of all the cms are not known for being long runners. Therefore I assume there might be some other problems. As well and by having a look on the other computers crunching the same work unit (#2551231), the work unit seems fine and finished in every case within less than two days.

Do you have rebooted or quit the BOINC client while these long running cms were running? All RNA World applications still have not got any kind of checkpointing and therefore the tasks will restart from zero once the BOINC client is restarted. A well working workaround is not to shut down the computer but using the hibernate mode instead. As well you should tick the option "leave task in memory while suspended".

One reason for this is, that we are using applications and algorithms which are designed for high performance computing and clusters, which do not have the need of checkpointing.
Hopefully there will be a solution for this problem soon. We know about some good approaches but have not got the time and man power to work on this. We will be happy about any help on this.

Concerning the parallelisation, it is still a big issue with BOINC. As far as I know Aqua@home is the only project using MPI within their applications. But there is now way of telling the BOINC client, that one task is using all available cores, so the BOINC client will start as many tasks as cores are available.
I know that there are some people working on implementing this into BOINC. But to be honest I doubt that this will happen within this year.

Cheers,
Torbjoern

PS: For us it is easier having the task ID or workunit ID given rather than the name. I searched for your user name, had a look on all computers attached to your account and saw your running tasks then.

[darwincollins]

I have aborted those 2 tasks. It is strange.
I don't think the checkpoints topic was relevent in this case. (its on a server)

[sqweeser]

Application: cmcalibrate v0.21
WU name: cmc_Lin64[Rfam-10.0-full]_RF00028_Intron_gpI_Rfam-10.0_1282002181_27_8
Result ID: 5468237

My loyal crunching computer with Xubuntu x86-64 crunched this WU happily little over two weeks and now when I came home from school, I noticed to my disappointment that Boinc manager and project page says it had computation error. I don't understand why. This computer has never had problems and boinc manager has a line:
Thu 16 Sep 2010 10:59:31 AM EEST RNA World Computation for task cmc_Lin64[Rfam-10.0-full]_RF00028_Intron_gpI_Rfam-10.0_1282002181_27_8 finished

and nothing about errors, so what's the problem? I hope I still get credit for this...

BTW, this was biggest WU I have ever crunched and these long running times and memory usage doesn't bother me at all as I can leave my computer on 24/7 (I said this just FYI). Memory usage was quite high in this WU, it topped to 2.74GB RAM in the middle of WU for some time and then dropped to ~1GB and stayed that level till end of WU.

Edit: I noticed that there actually is an error in the result page, I just didn't see it before: "process exited with code 195". So what exactly failed?