 |
|
Message boards : Number crunching : How can I limit climateprediction to run only one task at a time?
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 13 Nov 06 Posts: 2 Credit: 68,011 RAC: 0 |
I have a multiple core machine, but would like to limit climateprediction to only run one task at a time. How can I do that? Thanks! Phil |
 old_user544Send message Joined: 8 Aug 04 Posts: 26 Credit: 2,629,430 RAC: 0 |
From the \"Your Account\" page click on the \"View or edit\" for General preferences. Then on the new page that loads you will find a setting for \"On multiprocessors, use at most\". Set this to only one and the second core won\'t get a work unit next time. Mark |
|
Send message Joined: 1 Jan 07 Posts: 1066 Credit: 36,887,369 RAC: 1,533 |
It depends what you want to do with the other core(s). Follow md3\'s advice if you want to keep the other core(s) idle. But if you want to participate in one or more other BOINC projects (and provided your machine is powerful enough in other respects, such as having enough RAM, there\'s no reason not to),
The \'status\' column for CPDN should now confirm \"Won\'t get new tasks\": then you can re-edit your general preferences, allow BOINC to use the other core(s), and attach to other projects as you wish. |
|
old_user5994 Send message Joined: 31 Aug 04 Posts: 239 Credit: 2,933,299 RAC: 0 |
Les offered this advice ... Set to one core, allow CPDN to d/l the one task ... then set NNT as Richard noted, and set the cores to run as you desire. I have the same thought and suggested that CPDN add a preference to the site to allow us to have the scheduler be advised as to how many tasks we wish to DL, regardless of the number of cores. So far, to no avail. At least it seems that I am not the only one that would like this feature. In fact, I would bet that many are dismayed at getting 2 or more tasks that last for weeks/months and hate the thought of having them around that long, and hate eqaully the thought of killing the task for convenience sake... Anyway, hope this all helps ... I have 2, just down from 3, tasks on my 8 core and when I can run the CPDN model again, one should complete in another 3 full days of running ... then I will be back to the singleton I wanted in the first place ... |
mo.v Send message Joined: 29 Sep 04 Posts: 2363 Credit: 14,611,758 RAC: 0 |
The possibility of downloading just one task has been discussed on the boinc_dev forum and I think on the boinc_dev email lists, but I can\'t remember what if anything was decided. I once wanted a few SIMAP tasks to help align the finishing times of two CPDN models on my dual core. Within seconds about 75 SIMAP tasks flooded in and were off the bottom of the screen before I could stop more arriving. Each took 75 minutes to crunch, far more than what I needed. I\'d have liked an option to try just one first, and then select how many to accept afterwards. Cpdn news |
|
Send message Joined: 27 Jan 07 Posts: 301 Credit: 3,288,263 RAC: 26,370 |
The possibility of downloading just one task has been discussed on the boinc_dev forum and I think on the boinc_dev email lists, but I can\'t remember what if anything was decided. That\'s actually a hot topic (or maybe it should be?) with BOINC development. Computers attaching to a project should start with a quota of 1 task/day instead of the max. At least, that\'s how I would do it. I would have liked to see it in the 5.10.40 release, but as is the case with open source, not everyone agrees. |
|
Send message Joined: 13 Nov 06 Posts: 2 Credit: 68,011 RAC: 0 |
It depends what you want to do with the other core(s). Follow md3\'s advice if you want to keep the other core(s) idle. I DO want to run a few other projects. I DO NOT want the other core idle. I DO NOT want to limit a GENERAL preference. I DO want to limit ClimatePrediction.net preference to \"use at most 1.\" No wonder I was having trouble finding it... Is there any there there? Is there an Affinity that is some sort of analog for what I want? Thanks Phil |
|
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
The way that I do it, albeit just with the different types of climate models, is to set the 4 venues in general prefs for different numbers of processors, and the climateprediction prefs to different types of models. So, for 1 slab model, and 3 hi-res models: 1 processor / slab type. Upadate so BOINC on my computer knows about the setting. Allow BOINC to get work. Once the model is running, set for No new work. Change venues to a set that has 4 processors / hi-res type. Update again. Allow BOINC to get work again. Once the 3 hi-res models are running, set for No new work, and Suspend network activity, to stop BOINC telling the server about any failures, which in turn allows me to re-start form the most recent backup. ****************** For multi project setups, this may need to include setting all other projects to No new work first, other wise BOINC may grab something from one of those, instead of a climate model. And the \"get work ahead of time\" settings will also complicate matters. Backups: Here |
|
Send message Joined: 6 Oct 06 Posts: 204 Credit: 7,608,986 RAC: 0 |
Maybe you can juggle around with the resource share, where there is a will there is a way ;)I do it. Regards Masud. |
 Send message Joined: 3 Mar 06 Posts: 96 Credit: 353,185 RAC: 0 |
Juggling the resource share seems like it should work too but once the task is downloaded, if I understand the scheduler correctly, he should set the CPDN share to no more than 50%. If it\'s higher than 50% then eventually the other projects he is running will accumulate so much debt that the scheduler will stop downloading work for those projects which would leave 1 of his 2 cores without work. |
|
old_user5994 Send message Joined: 31 Aug 04 Posts: 239 Credit: 2,933,299 RAC: 0 |
It depends what you want to do with the other core(s). Follow md3\'s advice if you want to keep the other core(s) idle. THen the key is to get one model from CPDN., if you get two, kill one of them .... or suspend it. As for the test of it, then select resource shares as appropriate. Mine is more complex, but bear with me. I have 8 cores and insatiable desire to run projects. SO, I set resource shares from 12/14 through 100. I select up to 4 pojects with a share of 100 ... these will run continually on one core. In other words, I should see, in general, one task in work at all times for each project (variations in timing and oddities can cause slight bobbles and cases where this is not ALWAYS true, but it is rare). Should I want to put emphasis, I might have one at 100, and another at 300 where I would normally have 3/1 tasks running .... Four more projects get shares of 50, thus occuppying two other processors. so, half the time I should see work from one project and half the time the other ... Four more projects should get shares of 25 meaining ... I should see them rotate about a quarter of the time ... The last projects I give 6 share of 12 and two a share of 14 .... I see them once a day maybe slightly more ... And so, for example, I am trying to finish a CPDN task so CPDN is share 100 and runs all the time, as does M-Way and Malaria (225, so I have two and sometimes 3 in work) and SIMAP ... the rest of the projects get tailored shares ... as long as the number adds up to 800, I have a proper allocation of time ... With two cores, add to 200 ... though rememvber, with one task and a share of 200 for CPDN will not make it run on two CPUs. |
adrianxw Send message Joined: 31 Aug 04 Posts: 145 Credit: 2,080,724 RAC: 753 |
On my quads, I let it download some wu\'s, then set the project to \"No New Tasks\", then \"Suspend\" all but, (say), one of the wu\'s. When that wu is finished, I \"Resume\" another. So on until they are all gone, then I download some more. That way I can have however many CPDN wu\'s running as I like. With the real long deadlines, having a few non-started wu\'s lying around for a few months is no issue. Wave upon wave of demented avengers march cheerfully out of obscurity into the dream. |
|
Send message Joined: 1 Jan 07 Posts: 1066 Credit: 36,887,369 RAC: 1,533 |
On my quads, I let it download some wu\'s, then set the project to \"No New Tasks\", then \"Suspend\" all but, (say), one of the wu\'s. When that wu is finished, I \"Resume\" another. So on until they are all gone, then I download some more. It may not be an issue for you, but it may be for the project. If I remember correctly, if a CPDN server doesn\'t receive a trickle from a task it\'s issued every - I think - 60 days or so, it assumes that the computer has died, and sends out a duplicate task to another user. When I\'ve occasionally found myself in the position of having too many tasks in the past, I\'ve rotated them every week or so to keep the flow of trickles going up to the servers for all tasks. More recent advice, if you\'re doing that, is to avoid suspending a running task just after a trickle or restart dump (zip) upload - apparently the models are a little fragile at those points. |
©2025 cpdn.org
