| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: The Clean Energy Project Forum
Thread: Work Unit Pause: A and B type update |
| No member browsing this thread |
|
Thread Status: Locked Total posts in this thread: 79
|
|
| Author |
|
|
GIBA
Ace Cruncher Joined: Apr 25, 2005 Post Count: 5374 Status: Offline |
Hi everyone, So, as you all know, we are temporarily out of work units. The CEP team is working hard to get new good quality work units generated so that everyone can keep working towards CEP badges and new solar cell materials! Here is where we stand: A types: We are doing science. That means that some things don't work the way we thought they would ... but that means that we have a chance to learn from the errors that were returned. We are generating more A type work units, but we are also trying to minimize the chance that we will see the errors that have caused problems in the past. To do this, we have improved our "in-house" screening to try to get a better sense of which jobs will run smoothly and we are processing pre-work unit data this very minute. Again, this is only a pause - we have tens of thousands of molecules to test! B types: These jobs are a continuation of successful A type work units. We need more statistical data to analyze each computational job, so that is the role of the B type jobs. We are focusing on the work units that were completed correctly, so the jobs that caused serious problems will not be included in this round. To make sure things are running properly, these will be sent out in beta soon, so keep your eyes open! We are also working on a timing algorithm so that the WCG techs can more accurately estimate the time to completion. We really are working hard on getting all the kinks ironed out of this HUGE project - thanks for sticking with us!! The CEP Team Thank you very much for update us. I'm glad to see that this project are really evolving. I agree 100% with Rickjb (great job Rick !) and would like to highlight the part suggesting you guys from CEP take a deep look at DDDT2 project way of use CHARMM, once really appear be a very similar approachs between both projects. Just change information between projects could help CEP evolve quickly and almost without WU's errors. Additionally and maybe off topic... I would like to dare CEP team to think seriously about adopt GPU/CUDA for the next CEP project steps after this phase, once it could help you a lot of to speed up this project. Probably NVIDIA company team could help you on that, once have specifics to scientists that need run projects at high speeds with a oot of evaluations and math and so on (just try make a contact with NVIDIA company to discuss the possibilities and I now that you and his team will be surprised about...). It is just a single suggestion and obviously need be align with WCG team before once by now WCG don't support GPU/CUDA processing (but BOINC team support in the latest agent release for EISTEIN, SETI and GPU.GRID projects !). Just check it: http://www.nvidia.com/object/io_1229516081227.html http://www.nvidia.com/object/computational_chemistry.html  Edited: Just now saw another thread were CEP Team coment about GPU/CUDA...please disconsider my last coment about it: http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=24561   
Cheers ! GIB@
---------------------------------------- Join BRASIL - BRAZIL@GRID team and be very happy ! http://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=DF99KT5DN1 [Edit 2 times, last edit by GIBA at Feb 24, 2009 1:58:02 AM] |
||
|
|
kh6dc@arrl.net
Advanced Cruncher Joined: Dec 5, 2006 Post Count: 69 Status: Offline Project Badges: 
|
Mahalo Nui Loa (Hawaiian for Thank You Very Much)
 |
||
|
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges: 
|
I'm glad that my efforts seem to be (mostly) appreciated.
----------------------------------------ShadowJ, I was trained as a scientist, not a journalist, preacher, used car salesman or politician. When a scientist makes a statement, he has to back it up with evidence. If some or all of that evidence is a statement from someone else, he must provide a reference so the reader can check it out. That's what I have done (and it took me more than half an hour). I think it was in the Beta Testing Page Faults thread that one of the CAs (?) said that fixing the high number of PFs was not a priority (you dig out the link). In disagreeing, I had to make my argument as strong as possible, hence a more thorough approach. [Off-topic #1] Now for a suggestion regarding GPU computing (GPGPU, eg CUDA) on WCG, and essays. There seems to an ongoing trickle of enquiries about this, so perhaps it's time that someone wrote an essay about on why WCG should or should not use GPGPU, and had it put it into the FAQs. New enquiries on the subject could then be directed there. Include a list of pros and cons, and links to further info. Searching these fora for posts on the subject, and also Folding@Home would be good places to start digging. See also Wikipedia - GPGPU GIBA, the scientists & techs working on WCG projects are struggling to get the scientific modelling and distributed-computing software working correctly for CPU crunching. GPU crunching would be just too hard, and has other problems too. Since you are interested in the subject, perhaps you could start on the essay!  [Off-topic #2] Reminder: If you haven't done so already, don't forget to vote for the idea of Dell preinstalling WCG on its computers in the Dell Social Innovation Competition. Closes March 2. [Edit 3 times, last edit by Rickjb at Feb 24, 2009 1:01:20 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You're not wrong, Rickjb.
We would all love to solve the pagefault problem. But as ever, the techs have many other things to deal with, and resources are limited. CHARMM is closed source, and written in FORTRAN. Making even small changes will not be easy. GIBA: World Community Grid met with Nvidia already. You give them too little credit! :-) (This is why I get a little weary every time someone suggests GPUs as though nobody had ever thought of it before. We have known all about the possibilities (pros and cons) for years....) |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'm glad that my efforts seem to be (mostly) appreciated. ShadowJ, I was trained as a scientist, not a journalist, preacher, used car salesman or politician. When a scientist makes a statement, he has to back it up with evidence. If some or all of that evidence is a statement from someone else, he must provide a reference so the reader can check it out. That's what I have done (and it took me more than half an hour). I think it was in the Beta Testing Page Faults thread that one of the CAs (?) said that fixing the high number of PFs was not a priority (you dig out the link). In disagreeing, I had to make my argument as strong as possible, hence a more thorough approach. [Off-topic #1] Now for a suggestion regarding GPU computing (GPGPU, eg CUDA) on WCG, and essays. There seems to an ongoing trickle of enquiries about this, so perhaps it's time that someone wrote an essay about on why WCG should or should not use GPGPU, and had it put it into the FAQs. New enquiries on the subject could then be directed there. Include a list of pros and cons, and links to further info. Searching these fora for posts on the subject, and also Folding@Home would be good places to start digging. See also Wikipedia - GPGPU GIBA, the scientists & techs working on WCG projects are struggling to get the scientific modelling and distributed-computing software working correctly for CPU crunching. GPU crunching would be just too hard, and has other problems too. Since you are interested in the subject, perhaps you could start on the essay! [Off-topic #2] Reminder: If you haven't done so already, don't forget to vote for the idea of Dell preinstalling WCG on its computers in the Dell Social Innovation Competition. Closes March 2. Geez Rickjb, I was just joking with ya :P. I thought the fact I ragged on you for having too many quotes then got out a quote was an obvious enough contradiction to be noticed ^.^. Anyways, i'm rather tired of everyone acting like using GPUs is a new fantastic idea. It's great... if the grid had the spare resources. Notably, I say distributed storage and GPUs should still be a secondary priority, but the main priority at the moment should be getting more WUs and ironing out the kinks (errors). [Edit 1 times, last edit by Former Member at Feb 24, 2009 10:08:22 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks for the update!
 |
||
|
|||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Again, thanks for the update. And Rickjb, I hadn't realized that the problems with CEP were related to multicore processors. I work with a duo core and my system at home is quad core. I have plenty of disk space and I was wondering what I need to do to get more tasks in my que so I don't have a problem waiting for work; this is particularly true when my DSL line goes down or the internet stops working for other unknown reasons.
Regards, Dave |
||
|
|||
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges:
|
dkt: The problems that everyone's been having with CEP are not just related to multi-core CPUs. The only problem specific to multi-cores is the slowing of the system when 2 or more instances of CEP are running simultaneously. The main problems with CEP have been in getting the WUs to just run properly. I think that sometimes the atoms were bumping into an edge of the simulation space (check?). With others, the system energy would not converge to a minimum. Etc. The failure diagnostics and error handling could do with some tweaking. It's all due to a guy called Murphy I believe. The speed issue has been a lower priority.
----------------------------------------One other thing about CEP is that your CPU will run hotter than with any other current WCG project, and if you are overclocking or your CPU cooling is inadequate, you may find that your system is not as stable as you thought it was. In that case, you'll be unsure whether the error was your fault, or a bad WU. Re. tasks queues, there are 2 places that you can set these: a. In your "My Grid" >> "Device Manager" on the WCG website, click on the entry in the "Selected Profile" column in the row of the device (ie computer) that you want to set up; you will go to the page for that Profile, and among the Advanced Options are Workunit Cache Settings. BOINC will only read the change the next time it contacts WCG. b. In BOINC, there are "Local preferences", which override the WCG ones. In BOINC Advanced View, click on the "Advanced" heading and select "Preferences" from the drop-down menu. In the "Network usage" tab, set "Additional work buffer". Lots more info is in the Start Here: Community-maintained FAQs section, eg BOINC: Website Device Profile & Local Preferences Information. The Search button at the top of every forum page is there to be used, too. (Use the one near the middle of the row; it's more "advanced" than the one next to the box on the right). Happy crunching [Edit 3 times, last edit by Rickjb at Feb 26, 2009 6:46:57 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
RoaR... Rick on a Roll :O)
----------------------------------------Not noticed CEP causing extraneous heat, but 1 versus 2 certainly gave a 20% improvement. Now after a good while of testing using TThrottle to manage the laptop heat problem... 1.53, stable, valid results, smooth, not like the original BOINC CPU% time junket. http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=24557
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
GIBA
Ace Cruncher Joined: Apr 25, 2005 Post Count: 5374 Status: Offline |
You're not wrong, Rickjb. We would all love to solve the pagefault problem. But as ever, the techs have many other things to deal with, and resources are limited. CHARMM is closed source, and written in FORTRAN. Making even small changes will not be easy. GIBA: World Community Grid met with Nvidia already. You give them too little credit! :-) (This is why I get a little weary every time someone suggests GPUs as though nobody had ever thought of it before. We have known all about the possibilities (pros and cons) for years....) Didactylos, thank you to quote me with consistent content about it. Most of times, the answers received or saw about GPU/CUDA in forums here are just negatives or non sense discussions about barriers and issues. I'm glad to know that WCG talk with NVIDIA. It's a right way to start (and probably with ATI company too...) with GPU's. Hope that this conversation evolve asap and bring us good news, maybe for this project too. I'm a consistent user of GPU/CUDA in grid computing for almost a year, and can confirm that is great and useful crunch BOINC projects with this feature for instance in SETI, EINSTEIN and GPU.GRID projects. Folding@home is other of my favourites projects using GPU (in my opinion the best one...). It is important mention that in all project mentioned above, despite of some normal issues (similar with issues found by any early adopter of new technologies), the work with GPU's produces a lot of quick and good results !
Cheers ! GIB@
Join BRASIL - BRAZIL@GRID team and be very happy ! http://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=DF99KT5DN1 |
||
|
|||
|
|