 |
|
Questions and Answers :
Windows :
HadCM3 not allowing anything else to run?
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 7 Feb 08 Posts: 4 Credit: 214,397 RAC: 0 |
I\'ve had Boinc installed for several months, initially with just Seti@home, then later with climateprediction as well, configured 50/50. Over time, Seti seems to have been pushed out by cp such that now, Seti hasn\'t run in weeks or possibly months. I don\'t know much about how all this operates and reading the forums confuses me with all the nomenclature and idiomatic terminology, so please keep your answer simple. If I look in advanced view within the Boinc Manager, I see only 4 tasks running, all climateprediction. I can\'t seem to find a way to get Seti to run at all. Then I noticed that the 4 climateprediction tasks are all incredibly long tasks: hadcm3istd_0boc_1920_160_05936406_2 cpu time: 1919:25:11 progress: 81.252% To completion: 422:33:11 Report Deadline: 28/08/2010 8:37:22 AM Status: running, high priority The other 3 tasks are similar: cpu times: 1428:43:48/1312:38:38/1262:57:08 Progress: 59.056%/54.150%/52.515% To Completion: 884:01:42/976:50:59/1001:56:21 1919:25:41 hours equates to about 80 days of CPU, with another 17 to go. Is this normal? Seems awfully piggish for climateprediction to be working on tasks for 4 months solid. I really just want it only getting 50% of the available resources so that Seti can run. What do I need to do/check to make this work? |
|
Send message Joined: 30 Oct 05 Posts: 15 Credit: 96,265 RAC: 0 |
I\'ve had Boinc installed for several months, initially with just Seti@home, then later with climateprediction as well, configured 50/50. Over time, Seti seems to have been pushed out by cp such that now, Seti hasn\'t run in weeks or possibly months. Yes this is normal. BOINC thinks that it will not be able to complete the workunit by the deadline when running under your normal resource share. The models you are running are simulating 160 years of climate, so they do take a long time. Over the long term, BOINC will honour your resource share, as a debt is built up to SETI. When the climateprediction models have finished, SETI will run for the equivalent time it would have had, if the 50/50 split had always been used. Only after this debt is paid will climateprediction download new work. If you want to keep your resource share in the shorter term, you can switch to only running shorter models (this doesn\'t affect the ones you are currently working on). You can do this as follows Log into your account at http://climateapps2.oucs.ox.ac.uk/cpdnboinc/home.php Click on the climateprediction.net preferences option in the Preferences section. Under application preferences you will see the different types of model you can run. You will need to deselect the UK Met Office HADCM3 models, by clicking Edit climateprediction.net preferences at the bottom of the page, and then unchecking the relevant tick box. Make sure you update preferences when finished. Hope this helps Michael |
|
Send message Joined: 7 Feb 08 Posts: 4 Credit: 214,397 RAC: 0 |
Well, it\'s been 3 weeks since my original post. Since the, 2 of the 4 tasks have finished, and based on the response to my question, I expected that the climateprediction process would stand aside and allow the Seti to run. But that\'s not the case. When the 2 tasks finished, they were replaced by 2 more from climateprediction: This time, it\'s \"UK Met Office HADAM3 5.03\". So Climateprediction.net is still getting 100% of the share, and Seti hasn\'t run now in weeks. However, having said that, the log shows that Seti@home has had a lookin and done some work - albeit only short bursts of work lasting a few seconds, but there are many of them. Here\'s an extract of a typical set of log entries: 5/07/2008 8:11:11 AM|SETI@home|Finished upload of 28my08ab.25214.8661.5.8.42_0_0 5/07/2008 8:11:13 AM|SETI@home|Scheduler request succeeded: got 1 new tasks 5/07/2008 8:11:15 AM|SETI@home|Started download of 01my08af.13507.7843.14.8.32 5/07/2008 8:11:18 AM|SETI@home|Finished download of 01my08af.13507.7843.14.8.32 5/07/2008 8:11:19 AM|SETI@home|Starting 01my08af.13507.7843.14.8.32_1 5/07/2008 8:11:19 AM|SETI@home|Starting task 01my08af.13507.7843.14.8.32_1 using setiathome_enhanced version 527 5/07/2008 8:11:48 AM|SETI@home|Computation for task 01my08af.13507.7843.14.8.32_1 finished 5/07/2008 8:11:48 AM|SETI@home|Sending scheduler request: To fetch work. Requesting 30240 seconds of work, reporting 1 completed tasks 5/07/2008 8:11:50 AM|SETI@home|Started upload of 01my08af.13507.7843.14.8.32_1_0 5/07/2008 8:11:53 AM|SETI@home|Scheduler request succeeded: got 1 new tasks 5/07/2008 8:11:54 AM|SETI@home|Finished upload of 01my08af.13507.7843.14.8.32_1_0 5/07/2008 8:11:55 AM|SETI@home|Started download of 01my08af.13507.7843.14.8.142 Off the top of my head, it looks like Seti gets to run numerous times, but eventually Climateprediction wants to do some work and loads up with a gargantuan task that won\'t finish for weeks. All of Seti\'s tasks are rapid and short-lived. The end result is that 99.999% of the time, it\'s Climateprediction that\'s running, and over the last 2 months, Seti\'s received almost no time at all. Doesn\'t seem fair. At what point does the climateprediction new task requests get told \"bugger off, you\'ve hogged the CPU for the last 2 months and you won\'t get another chance until Seti\'s hogged the CPU for an equivalent amount of time\"? It seems that the rule is more like, \"Bugger off climateprediction, let\'s give Seti a chance. Ok, Seti, go ahead. (30 seconds later...) Ok, Seti you\'ve finished, let\'s give climateprediction another go\", at which point, CP loads up a 2-month monster task yet again. |
|
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
1) In your BOINC manager under the Projects tab, click the cpdn project to select it, and then click No new tasks. 2) On your account page on the server, look in climateprediction.net preferences. There are sections for the 3 types of model. Click the type that you want: A) HADSM3 (Slab) - short and low memory. B) HADAM3 (Hi Res) - also short, but high memory requirement. C) HADCM3 (Coupled Ocean) - low memory, but very log. (Actually 2 types: 80 year models, and 160 year models.) As for climate models hogging your computer, BOINC is just doing what it\'s been designed to do - allocate cpu time in the proportion requested over the projects that you\'ve chosen. When people add a climate model to other short work units, the time period for \"equal time share\" becomes spread out over months, rather than the hours or days that it would be if only short WUs were to be run. This is the way it is. Backups: Here |
 mo.v Send message Joined: 29 Sep 04 Posts: 2363 Credit: 14,611,758 RAC: 0 |
Hi Dir I don\'t know much about SETI, but I\'ve looked at a handful of your SETI results, for example this one. I\'ve read on the SETI forum that about 5% of SETI tasks end very quickly with this message about the number of results exceeding the space allocated. All your tasks I looked at had finished this way, though occasionally somebody else\'s computer has crunched a task from the same workunit to the end ie their computer\'s task from the same WU hasn\'t ended prematurely. I don\'t know what this message means, but it seems a bit of a coincidence that so many of your SETI tasks are finishing prematurely with the same message. (I don\'t think it means that the number of aliens/alien communications you\'ve detected is more than we have space for on planet Earth.) What Les has said about the way the BOINC scheduler works explains a lot, but I\'m wondering whether your computer has a problem with its SETI tasks. Rather than ask you to take this query to the SETI forum, let\'s see first whether Richard Haselgrove (for example) can tell us here whether your crunching for SETI is normal & successful or not. Cpdn news |
|
JIM Send message Joined: 31 Dec 07 Posts: 1152 Credit: 22,098,944 RAC: 3,014 |
Dear dir If you want to run Seti for a while there is a simple solution. Simply suspend the Climateprediction models for a few days at a time. To do this: First, as Les advised, be sure “no new tasks†is clicked in “Projects†to be sure that you don’t get any new WUs. Then click “Tasks†and highlight the climateprediction WUs. Then just click “Suspend†for each WU. Any Seti WUs that you have downloaded should run. When you want to run climateprediction again just click “tasks†and highlight the WUs and click resume. You could do this every 2 days or so. I realize that this is a bit of a pain, but, it will get crunching Seti again. Well, it\'s been 3 weeks since my original post. Since the, 2 of the 4 tasks have finished, and based on the response to my question, I expected that the climateprediction process would stand aside and allow the Seti to run. But that\'s not the case. When the 2 tasks finished, they were replaced by 2 more from climateprediction: This time, it\'s \"UK Met Office HADAM3 5.03\". |
|
Send message Joined: 1 Jan 07 Posts: 943 Credit: 34,312,322 RAC: 11,337 |
Hi Dir It\'s a good thing I dropped by to wish happy landings to 7327356.... SETI has had database problems over the weekend, which has reduced the amount of available work, but that doesn\'t explain your situation. As Mo has noted, every one of the SETI tasks visible on your account pages has ended with SETI@Home Informational message -9 result_overflow SETI is looking for radio signals from outer space. Every so often, local (earthly) radio signals get caught up in the recordings - anything from tourists\' cellphones to air traffic control radar. So SETI has a mechanism for ditching contaminated recordings, so you can move on and process something significant. The result_overflow message signifies this process. But it should only happen to a small proportion of tasks, and all the computers allocated that task should agree that the recording is contaminated. In your case, 100% of the tasks are reported as contaminated, but your co-crunchers don\'t see the same contamination. There is a known, rare, bug in the SETI application which causes a false contamination report (it\'s been discussed at SETI Beta), but that only generates a report of multiple (31) pulses - your tasks also show spikes, triplets and gaussians. So I can only conclude that the reported \"contamination\" in your results originates in your own computer. Until you find and correct that problem, there\'s no point doing any work for SETI - your work has no scientific value, and you are getting no credit as a result. There is also something strange about your Climate tasks. Yes, they take three months or more on a Q6600 - like the one which I have reported while typing this - but that shouldn\'t trigger \"high priority\" in BOINC: the 3 months computation can be spread out over 30 months to deadline, so even 3 hours per day would be enough to keep BOINC happy. I suggest you work with the experienced crunchers - both here and at SETI - to work out what is going wrong with your computer. Please tell us a little more about it - motherboard, RAM (we know the CPU), whether it is overclocked, what other programs you run, that sort of thing. There\'s something a bit odd going on here, but without more information, I can\'t quite put my finger on it. |
|
Send message Joined: 7 Feb 08 Posts: 4 Credit: 214,397 RAC: 0 |
Ok, will dump some details about the environment, then work out who/how to get that info to the right persons. The PC was new (about 3 weeks old) when I installed BOINC and it\'s been running 24x7 since, in an office environment. I only do email on it and Word, so there\'s nothing computationally intensive happening. The PC is a managed corporate machine. I have local administrator access when I need it, but I normally run in a restrictive user logon account. I installed BOINC as administrator and it runs using the administrator account as a service. I log on as a restricted domain user and run BOINC Manager from there. The PC has several corporate things installed: Lotus Notes 6.53P Lotus Sametime 7.5 MS Office Citrix Desktop (but I\'m not running in a Citrix env) Quicktime, VLC, RealOne, Windows Media player Zenworks for Desktops 4 etrust A/V I\'m running an Acer Veriton 7900Pro, which is an Intel Core 2 Quad Q6600 (Kentsfield)@2.4GHz. Not overclocked. GPU is an Nvidia GeForce 7600GT, not overclocked, 256MB, on ForceWare version 175.16. I have some sort of graphics issue that I haven\'t isolated yet - about 80% of the time, images appear corrupt onscreen. It appears the same as when you overclock a GPU and you start getting artifacts. I live with it but it\'s annoying. Opsys is XP SP2, DirectX 9.0c. Boinc Manager is 5.10.45. Someone let me know if there\'s any other info you\'d like, and if I need to contact anyone else to look at this. Cheers, Doug |
|
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
I have some sort of graphics issue that I haven\'t isolated yet - about 80% of the time, images appear corrupt onscreen. If you\'re talking only about climate model images, there\'s a known problem with slab models, (which at present, are the only ones with several phases), where the text characters become blobs at the end of each phase. This can only be cured by Exiting from BOINC, and then re-starting it. If you haven\'t been doing this, then for 2/3rds of the time the text will \'be corrupted\'. Backups: Here |
|
Send message Joined: 7 Feb 08 Posts: 4 Credit: 214,397 RAC: 0 |
I have some sort of graphics issue that I haven\'t isolated yet - about 80% of the time, images appear corrupt onscreen. No, I don\'t use any graphical interface for BOINC stuff. |
©2024 climateprediction.net
