 |
|
|
41)
Message boards :
Number crunching :
[Wish] Show realtime crunching speed in the screensaver?
(Message 20016)
Posted 7 Feb 2006 by  old_user5994
Post: If the work is in progress the system will slowly correct. I upgraded the speed of one of my systems and it WAS 3.79 sec/TS ... at the moment it is 2.94 and still falling hundreth or so a day. My hope is that it will get down to 2.79 or less. It is a dual core chip, but the basic performance on other work is roughly equivelent to another of my systems, a single core 3500, and that is about what that system does. I don\'t expect that the time will fall that fast now, and that it will take longer to lower the time average. The truth is that none of these systems really display instentaneous performance anyway. This is made worse by the fact that for more normal work the majority of the projects work now is highly variable in time length. About the only project where teh work seems to always be roughly the same size is SIMAP ... SETI varies by angle range (soon to be worse), Einstein by something (I don\'t know for sure, but I have 50 min to 4 hours on my fastest system), LHC has early death, ROsetta has variance because the work varies. |
|
42)
Message boards :
Number crunching :
Sulphur units constantly failing
(Message 19985)
Posted 5 Feb 2006 by old_user5994
Post: Guys, I am not sure if this is completely relevant, but, I had two idle computers and downloaded work on both. One has 12 hours in on the work, the other died immediately. Looking at the ones that are dying right away all of them have been recently created. The one that started running was generated back in August. I mean, it may die soon too ... but, are we sure that the work being created now is valid? This is really odd to me, computers that just recently were successful in doing models now cannot even start one up? |
|
43)
Message boards :
Number crunching :
Sulphur units constantly failing
(Message 19975)
Posted 4 Feb 2006 by old_user5994
Post: It jsut occurred to me that this would HAVE to be a BOINC Client Software change as along with the CPDN slot folder back up, a client state file extract would also have to be made. Requiring the participant to back up the entire BOINC folder is obviously impractical for those of us that run multiple projects as ther remaining project contents would be very dynamic. I am not sure how the system as a whole will react with a client error that has already been reported if such a back-up system was in place in that the \"rewind\" would \"reserrect\" the dead! :) |
|
44)
Message boards :
Number crunching :
Sulphur units constantly failing
(Message 19972)
Posted 4 Feb 2006 by old_user5994
Post: Frequent backups are a must! Well, if this is true, then it needs to be part of the application. Like Clark said, it is difficult for me to remember/perform daily back-up on the CPDN project on the Daily even times ~80 days times 8 systems is a lot of additional work. To this point I have had models die on occasion ... but, on systems that had completed regularly and without trouble 4.19 models ... My only concern is that I am wasting my and your time even trying to run these models. I cannot be sure, but I do not think I have yet to complete a 4.22 model. Many die immedately others seem to wait a bit .. Ok, I saw on the other board that some new work has be created so I will try a couple more. ==== edit I don\'t mind the space, and would like to have \"rotations\" (for me I would pick 2) but there is litterally no way that I am going to remember to do this consistently enough to be of value. In my situation it is just too hard for *ME* to do on a manual basis. Perhaps this should become a fairly high priority item for the Devs on the next major update when the new models are released. In all honesty, it probably should be made part of the BOINC application so that any project could avail itself of the utility. The control would be on the Preferences page for the project, allow backup or perhaps simply a number daily back up allowed 0, 1, 2 to probably a max of 4 ... |
|
45)
Message boards :
Number crunching :
Sulphur units constantly failing
(Message 19967)
Posted 4 Feb 2006 by old_user5994
Post: I just had another 4.22 die on me at 60% (?) forgot to write it down, it was at either 40 or 60%, It was this one. I know it says aborted by GUI, but the app was not running, the directory had only files that were 1K in size (I did save this as a copy, send me a note if you want the copy of the slots dir). Very strange indeed. the only good news I guess is that I still have 3 4.19 models and they seem to be runing well. The tension is rising ... 1 hour something on one of them .. :) I hate to be brusk, but are any of the 4.22 models completing? And, even it not, is it worth my time to run paritals? I know you probably told me somewhere ... Anyhow, p.d.buck@comcast.net if you want the slots dir ... not sure what good it will do, all the files as small and zips ... a runaway delete all files gremelin? NO matter I suppose ... :) |
|
46)
Message boards :
Number crunching :
WUs constantly failing
(Message 19913)
Posted 2 Feb 2006 by old_user5994
Post: Perhaps for your friend a different project might be more appropriate. I know this sounds like heresy ... but, not all project are suitable for all computers and all people. I have had decent luck running CPDN on all my PCs, occassional model crashes for various reasong, but, aa pretty decent track record. Heck I am about to complete my second Sulfur model in a coupld days (1 day 12 hours). But, though you would think that it would be a better computer to run CPDN I have yet to complete a model on my PowerMac G5 ... bad computer? Bad program, gremlins? who knows. But, I just stopped and now run other projects on the PowerMac, it really shines at Einstein@Home ... Again, this is the beauty of BOINC ... Oh, and WCG uses the UD program if you like, or you can run thier two projects under BOINC like I do ... |
|
47)
Message boards :
Number crunching :
WUs constantly failing
(Message 19904)
Posted 2 Feb 2006 by old_user5994
Post: Well, I wish mine were that simple. Stable, single use platform, BOINC only ... heck, it is so single use I usually only look at it through RealVNC as there is no need to go local ... :) |
|
48)
Message boards :
Number crunching :
WUs constantly failing
(Message 19891)
Posted 2 Feb 2006 by old_user5994
Post: It did not take long ... error and it failed for the other participant on start up ... Are we sure we are done with the bad work units pending? Well tomorrow is another day. I looked in my account, I thought I had done more sulfur, but so far have only successfully completed one. But, I do have another that is only 2 days from completion and it runs continuously, so, theory says that should be a good estimate (though I am not sure ... probably will take 3-4 days). Regardless, I have one more coming, the next one after that has 16 days to run ... |
|
49)
Message boards :
Number crunching :
WUs constantly failing
(Message 19889)
Posted 2 Feb 2006 by old_user5994
Post: Well, I have been doing pretty well I thought. But one computer now seems to have had several in a row. I am going to try to get one more to see ... I had been completing and had completed several SLab models with that computer and the first of the three deaths seems to be the cross from phase one to two ... I think that was mentioned as an issue. The last two were \"819\" traps on start up. Which I find interesting as I don\'t run the graphics and that error trap is USUALLY an indication of a video card issue ... well I will try one more I guess ... ==== edit Forgot to mention that soneone else also tried the last two models and also got an \"819\" error, though one DID run for a bit ... interesting ... Really strange ... |
|
50)
Message boards :
Number crunching :
Lighthouse for novice Mac volunteers (interim)
(Message 19885)
Posted 2 Feb 2006 by old_user5994
Post: Neither does WCG or SIMAP@Home at the moment. Both have stated that they are planning on them ... but ... |
|
51)
Message boards :
Number crunching :
Hypothetical? funding plan
(Message 19866)
Posted 1 Feb 2006 by old_user5994
Post: See my notes in the other Forum ... But, I like what you did. A clean start and simple explanation... Ensemble is good to expand on the term as we should. Now we need to link into the other pieces ... Also, for those looking on ... I am restructuring the glossaries ... so we will have the one big one, plus smaller ones ... will it make things easier to find? I don\'t know, but, all I can do is to provide as many ways to \"slice\" the pie as I can and hope it helps ... |
|
52)
Message boards :
Number crunching :
WUs constantly failing
(Message 19852)
Posted 1 Feb 2006 by old_user5994
Post: Me too ... I just killed two .. don\'t feel bad, i notice that they both failed for someone else too ... |
|
53)
Message boards :
Number crunching :
Hypothetical? funding plan
(Message 19845)
Posted 1 Feb 2006 by old_user5994
Post: SOrt of the idea of what I had in mind is something along the order of this: An Introduction To The Science Behind Climateprediction.net. Most of the material currently \"jumps\" right into the models and how they all are calculating the salinity of the water ... The proposed article is obviously NOT COMPLETE! But, it illustrates what I meant I think. I stuck it into the top of the template of the models so it will show up wherever the template is used and that also may not be what you guys would like. Most importantly, I do not know if the stuff I said is even close to being right. But, it is more or less what *I* gather. But, a simple, non-technical start, with a gradual increase in the technical content. What else should be there would include a \"work flow\" showing a model issue, the participants returns, how they are aggregated and evaluated, and then followed by the creation and issue of new models in the series. Followed by the modification to the model structure, like slab to sulfur, and so on ... Can you see what I mean? The articles being written are nice, but they are all about the in-depth science of the infividual test cycle, but not what the heck the project is DOING ... :) Once we get past the introduction, THEN we lead into the details of each of the expiriments. Which writeups by the way are still missing what the heck we learned from the models and how we learned it. For example, the Slab models are done. Ok, cool, what did we learn? How did we learn it. If this is in published papers, then the cool thing would be to list the papers, AND to extract salient points in the Wiki as a summary. For example, I remember seeing a graph with lines that curved up and to the right showing the basic divergeence of the models. Where that chart is I could not tell you, but, i recall that the explanation did not help me to understand the connection of what *I* did to that chart ... I looked at the cost of the OC school and it looks to be over $200 US the only way I can find out for sure is to sign up and pay for it I guess ... but, again, all I want to do is to read the material, I don\'t want to get a degree (I already have enough for now, most disabled people don\'t need more than a masters ... :)). |
|
54)
Message boards :
Number crunching :
Hypothetical? funding plan
(Message 19823)
Posted 31 Jan 2006 by old_user5994
Post: Interesting post, Paul. Is this what you were talking about on the other forum, when you said you wanted each project to document BOINC? Yes it was. OU course ... my guess is that I would need to come to Oxford to attend? A big problem for those of us that have, ahem, issues with airplanes ... :) I will look at those links and see if I can \"shoehorn\" them in at least a little bit. Some how. Anyway, this is certainly why I feel the Wiki is the way to go. Each of us can contribute just a little. Sometimes when I get \"just enough\" information I can start to \"fill in\" pieces that are not obviously missing to those that have more (too much?) experience in some things. It is one of the reasons it is so hard to write... and why I have been unable to do as much lately as I would like. Doing good writing is extrodinarily hard. And now I have something like four articles to re-write ... :) |
|
55)
Questions and Answers :
Windows :
Unrecoverable error for result sulphur_iy78_000884132_0
(Message 19811)
Posted 31 Jan 2006 by old_user5994
Post: Michael SETI@Home, Einstein@Home, Rosetta@Home ... The real kicker is that on some systems the problem is easily found in that it happens on screen saver use. Cure, no screen saver. However, in MOST cases, the program causing the error seems to be external to BOICN. Which programs, or if it is simply bad video drivers, is still unknown. Again, some have solved the issue with updating the video drivers. Others have not. If you can locate the sensitive area it would help ... like does it happen when you launch Quake with an ATI card? Or Doom IV with nVidia? |
|
56)
Message boards :
Number crunching :
Hypothetical? funding plan
(Message 19810)
Posted 31 Jan 2006 by old_user5994
Post: Chris, Aside from personal debt which precludes my contributions to a project. The point I was trying to make is that the visibility of what is gained just really is not there. I donate based on the hope that the CPU cycles do mean something. But, all I can say with assurance is that I have completed \"x\" models, failed at \"y\". There are scientists looking at the \"stuff\", but, what directly is the result of *MY* contribution is invisible. I think I have said this before. If you do NOT come to this question with a completely blank mind you probably cannot see my point. In other words, if you know where the data is and what it means, well, then you will have a hard time realizing what I am trying to say. Just as an example, which I was going to write to you about, is the data related to the models that is captured on the Work/Result pages ... what does that tell me? It is not explained anywhere. At one point, I had an explanation page for each and every form and page in the web site. But, since I have been unable to convince any project, or Dr. Anderson of the utilitiy of having a \"Help\" page for each page in the web site the usage numbers of those pages, um, stunk... an example. The data I referring to are the entries in the table \"Perturbed Parameters for Result # 1634604 Sulphur Cycle Experiment\" and the two links below. But, this is a symptom of the question you are asking. What does it mean to the participant? All I know is that I have done \"x\" models, but do not really have a clear idea of what those models mean, how to interpret the data, how you combine the data from \"n\" models into a something else ... This is back to the fundamental issue of technical writing. How do you explain something to someone with little or know background? Where do you start? The Wiki now, thanks to your work, has a nice intro on getting started. More data on the basic science you are trying to do, and nothing that ties the two ends together. Rosetta@Home showed some energy models returned by the participants. But, have been failing to update those as expected ... ahem ... something I have also been a long critic of for almost all projects.... :) With regard to direct funding, I probably would donate some to most of the projects, and/or BOINC with the only reservation that I have no idea what the funding would be used for... if you were to launch a subscription drive for a widget that will assist the project you might be more successful. I for one, have no idea what the project would find most useful. What would be most useful to me is the trail of bread crumbs from my returned results through the process you go through to the conclusions ... Anyway, this is the same point I tried to explain in the past ... what does what I am doing mean ... is my model proving that the globe is warming, is it creating showstorms in the sahara, is it a complete waste of time? How the heck do I know ... SETI@Home has collectted some amount of funding from contributions, some hardware donations (I believe) from participants, but, where did it go? What impovement did it make for the project? Dave, good numbers, goodd rebuttal for my contention that ya\'ll don\'t communicate enough ... with one or two possible objections ... :) how much of that 10% was talking back and forth between staffers, and two, for those of us that don\'t visit those forums it is like a UK telephone call ... still un-tapped ... again, *I* don\'t go there as I can see what threads I have not seen all the posts, but within the thread I have to read the whole thing again, I can\'t have it sort to my preference on entry, etc .... it is possible these things can be done, but I have not found the magic yet. Which is why if it is not here in this forum I don\'t see it at all ... and with BOINC the delivery platform, this is going to increase in importance ... but we had this conversation too ... :) Does this help your question Chris? |
|
57)
Message boards :
Number crunching :
Hypothetical? funding plan
(Message 19760)
Posted 29 Jan 2006 by old_user5994
Post: SETI@Home already has this type or thing set up. Though, they do not have the ability of the participants to direct the use of the funds. I too am into this for the science. If you look closely at my signature you will note that the tendency is to put the most effort into those projects that are actively doing science ... well, if you look at absolute totals it is not quite so clear because SETI@Home had nearly a year\'s running start ... but, if you you look at the totals and the RACs you can see that the effort is aimed towards CPDN, Einstein@Home, SIMAP, WCG right now ... with the rest getting some low amounts, Rosetta@Home because the work there will help WCG later, LHC@Home because we need the LHC, and we may start to get actual science work there in the future. But, at most projects we have low visibility into what all our work means. I have stated this before in other forums and at other times, but, the results go in and we really don\'t get to see much of what it all means. At least not in a form that I think is usable to an average person. The science papers are fine, but we need more writing like the article done for SETI@Home \"how it works\" etc. So, to my mind, I would like to see more spent in that direction, would that take more servers? More people ... another funded position? Could this be used in that way even? |
|
58)
Message boards :
Number crunching :
WUs constantly failing
(Message 19561)
Posted 23 Jan 2006 by old_user5994
Post: Um, Memtest86 is for a PC ... not a G5 ... |
|
59)
Message boards :
Number crunching :
Sulphur model phase 3 unchanged since phase 2
(Message 19497)
Posted 21 Jan 2006 by old_user5994
Post: I had one where all of a sudden the s/TS was super low ... I just let it run and eventually it died. One of the \"fun\" things in CPDN, and Rosetta@Home is that there is almost no such thing as a \"failure\". The models show something even when they go \"bad\". The fact that we get credited for the effort doesn\'t hurt ... after all, it is all about the credits! :) |
|
60)
Message boards :
Number crunching :
Wiki Navigation
(Message 19442)
Posted 19 Jan 2006 by old_user5994
Post: Chris R. from CPDN asked if we could modify the navigation side-bar for the Wiki, I think we have a \"stable\" version of the concept. Take a look and let me know what you think ... |
©2024 climateprediction.net