Posts by Dagorath

1) Message boards : Number crunching : Exceeding maxium disk space. (Message 21303) Posted 11 Mar 2009 by (banished: ID 70524) Post: Gas Giant wrote: tpieloni wrote: Yes, I am from CERN/LHC and I am new with LHC@home... I am running some studies for the LHC and trying to learning how to use this powerful pool of CPUs. I will try to stop by the forum every now and then to update you on the outcomes of the study. Let\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\'s keep running. Excellent. Welcome from us all! tpieloni is one Gas Giants\\\' numerous sockpuppets
2) Message boards : Number crunching : Exceeding maxium disk space. (Message 21272) Posted 26 Feb 2009 by (banished: ID 70524) Post: My Linux box, host 9780021, returned a Success on this one.
3) Message boards : Number crunching : Exceeding maxium disk space. (Message 21268) Posted 26 Feb 2009 by (banished: ID 70524) Post: Are we seeing another HR scenario where the same science app behaves differently on different hosts? I think Eric Macintosh licked that problem a long time ago with new floating point libs. My guess is that given the same work unit, Vista under-reports the disk space the app is using. Or maybe XP and Linux over-report the disk space used by the app. Either way, if the WU simply sets the max a little higher there should be no problem. Just guessing, I have no evidence to suggest Vista over-reports or that Linux/XP under-report.
4) Message boards : Number crunching : Exceeding maxium disk space. (Message 21254) Posted 25 Feb 2009 by (banished: ID 70524) Post: So what could nI do ? Is it the new Sixtrack or is it the new BOINC version ? Thanks for quick help Dr.Mabuse It is not a problem with your settings. There is nothing you can do. Every work unit from every project has a maximum disk space number. That number is determined by the project, not by BOINC client. In this case it looks like the maximum is 28.61MB. It is just one more safety mechanism. LHC is telling your BOINC client that the task\'s output files should never use more than 28.61 MB. Either BOINC or the Sixtrack application found the task was using 38.99MB so it aborted the task. Either the new Sixtrack application has a bug which causes the output files to grow beyond the maximum or else they are underestimating the maximum size of the output files.
5) Message boards : Number crunching : Why BOINC 6.X has issues with LHC@home and other things (Message 20838) Posted 29 Nov 2008 by (banished: ID 70524) Post: Pay at the first window. Would you like to upsize your order to 1500 tasks for just $12.99?
6) Message boards : Number crunching : WU \\\"Aborted by Project\\\" (Message 20835) Posted 28 Nov 2008 by (banished: ID 70524) Post: ...it sends a message to your Computer to abort the WU. That could be taken 2 different ways. It doesn\'t really \"send a mesage\" in the same sense that you send a letter. In other words it doesn\'t initiate a connection with your computer. It waits until your computer contacts the server, then the cancel message is delivered. Very often your computer does not contact the server in time to receive a cancel message so it goes ahead and crunches a task that doesn\'t need to be crunched. That is why using the cancel mechanism is not a good way to eliminate the redundant work and waste inherent in the IR > minQ strategy.
7) Message boards : Number crunching : Why BOINC 6.X has issues with LHC@home and other things (Message 20834) Posted 28 Nov 2008 by (banished: ID 70524) Post: Donations are a good idea, FalconFly. Here\'s another... Since so many crunchers here think it\'s just the most wonderful thing in the world to be able to crunch LHC@home tasks, they won\'t mind paying for such an incredibly fulfilling privilege. Sell the work units for 1 penny per task. $10 would get you 1000 tasks.
8) Message boards : Number crunching : Why BOINC 6.X has issues with LHC@home and other things (Message 20796) Posted 22 Nov 2008 by (banished: ID 70524) Post: YAY WORK! Can someone explain why I have on each machine, a bunch of WUs that finish properly, and a large amount that are tagged as redundant result, or cancelled by server. I did a search in the message boards, but got no results. Is that because the search doesn\'t work either? Does this have anything to do with the BOINC 6.x problem you are all talking about? Cragg They send out 5 copies of each task, each copy goes to a different computer. Consider those 5 to be an ad hoc temporary work group. The other 4 member are your wingmen for the mission. 5 is the Initial Replication or IR for short. They need only 3 matching results to verify that they have \"the right answer\". 3 is the minQ, short for Minimum Quorum. If they get 3 that match before you start crunching your copy, they cancel yours because it is not needed (redundant). It would be a waste of your CPU time if they didn\'t cancel it. It has nothing to do with the BOINC 6.x problem directly. The need to send out more than 3 copies is a direct result of the fact that this project has always, even before the BOINC 6.X problem, had a high rate of compute errors and results that fail to match the other results returned by members of the work group. The cancel mechanism does not always work because the cancel order frequently does not reach your computer in time. Then you waste CPU time doing work that doesn\'t need to be done. It is far more efficient (less wasteful) to send only 3 tasks (IR = minQ) and wait until those all return and then, if the 3 don\'t match, send a fourth copy. The reason they use IR > minQ is because they know from experience that some 25% of reults will fail to match. Thus IR > minQ speeds up the completion of the batch but at the expense of precious CPU time. There are other strategies for speeding up the completion of the batch that are much more efficient. The IR > minQ strategy is an anachronism from long ago. The Advanced Search works much better but it\'s far from perfect.
9) Message boards : Number crunching : Why BOINC 6.X has issues with LHC@home and other things (Message 20784) Posted 18 Nov 2008 by (banished: ID 70524) Post: Nice try but guess again. So, where in the world can you divert funds from a gov\'t funded research project by simply \"[speaking] to the right person\"? Just about everywhere in the world, in every public and private enterprise. It\'s simply bad management to lockdown every penny and require \"the board\" to vote on every penny spent. Every enterprise has at least 1 person who has the authority to divert funds and spend small amounts to meet unforseen circumstances. It\'s absolutely necessary to avert disaster and/or keep the enterprise on an even keel. Fools and their money always part quickly. I\'m gladly giving less than $0.1 a month to what I consider the most impressive project since the great pyramids were built. If you think that\'s foolish, fine for you, I couldn\'t care less. Personally I think it\'s great that I can contribute, even if it\'s just a tiny tiny bit. It IS the most impressive project since the great pyramids, don\'t think even for an instant that I don\'t appreciate what a marvel of engineering it is and how badly we need the knowledge the collider will provide. That doesn\'t mean we should piss away CPU cycles needlessly just to make the collider work. The fact is you could contribute even more if they were not wasting so much. Hey, it\'s not like the waste is unavoidable. I don\'t whine about waste that is unavoidable. But the waste here IS avoidable to a large measure. Now about the money... OF COURSE I don\'t worry about the $.1 you waste every month. It\'s always been about the TOTAL waste in the project. And it\'s about the TOTAL wasted CPU cycles, not really the wasted money. Yeeesh! I thought you were reading my posts and on the same page.
10) Message boards : Number crunching : Why BOINC 6.X has issues with LHC@home and other things (Message 20779) Posted 18 Nov 2008 by (banished: ID 70524) Post: Regarding diverting funds... there is someone who can divert funds, there always is. If funds haven\\\'t been diverted then you just haven\\\'t spoken to the right person or else you have but they\\\'re pretending to be deaf. In other words, you have absolutely no idea what you\\\'re talking about at all. Nice try but guess again. As for me, I consider the scientists time more valueable than my CPU\\\'s idle cycles. Fools and their money always part quickly.

Previous 20 · Next 20

LHC@home