Posts by Ross Morgan

1) Message boards : Number crunching : Report Deadline (Message 11328) Posted 11 Nov 2005 by Ross Morgan Post: Ok, so I am new to LHC and have probably missed this, but why does the newest result I have received have a reporting deadline sooner than the one before? The LHC scheduler has an optimiser built in that estimates the best deadline to get the studies done in the shortest overall time. The optimiser calculates deadlines on the fly based on return times, the number of computers available, and the size of the study (amongst other things). It is quite possibe that the deadlines have shortened between downloads such that later units finish earlier than earlier ones. That's all well and good however i am getting 4 and 5 day deadines to complete WU's and the scheduler is downloading far too many WU's to get them completed in time. The LHC@home scheduler does not seem to take in account the LHC@home project share of cpu time or the fact that I have 4 other Boinc projects running on my PC. It looks like LHC@home will be getting suspended on my PC until we return to something even approaching a sensible deadline for complettion.
2) Message boards : Number crunching : Attention LHC Developers! (Message 10814) Posted 20 Oct 2005 by Ross Morgan Post: <blockquote>Indian Summer in Denmark has broken. Raining today, outlook for the next 7 days, wet, windy, typical Denmark weather in fact. Had our first frosts this week as well, more forecast. Was good while it lasted. Will be good to have v5 support. I only run the versions I do now because of LHC. The fixes are becoming mandatory at some of the projects, so a crunch point is not far away.</blockquote> You think Denmark's weather is bad? You should try living in Scotland for a while. :-d
3) Message boards : Number crunching : Attention LHC Developers! (Message 10796) Posted 20 Oct 2005 by Ross Morgan Post: Surely it can't be that hard to patch the servers to accept the V5 clients. Is it a big and complicated job to upgrade or something? all the other Boinc projects upgraded ages ago.
4) Message boards : Number crunching : Reason for not getting work: won't finish in time... (Message 10641) Posted 8 Oct 2005 by Ross Morgan Post: <blockquote><blockquote><blockquote>Ross, Post your boinc-project allocations, boinc client uptime %, and boinc client version we can tell you why it's doing that. It took my 10 active hosts about 1.5 months to finally even out to something which resembled predictability. The Boinc client does work correctly and in 99%+ of cases it's right. </blockquote> My project allocaions are:- Einstein 300 (42.86%) Climate Prediction 100 (14.29%)(Currently suspended) Predictor 14.29% 100 (14.29%) LHC@Home 100 (14.29%) Seti 100 (14.29%) Client uptime 63.8777% Boinc Client Version 4.45 </blockquote> If the next project is Einstein, then there is a factor of 3, since the uptime is only 63% (and on time is about 50%) this is probably about another factor of 3. This gives a factor of 9 that the time remaining is multiplied by to determine if any more work should be downloaded now. So a few hours (say 10 for conveniece) is suddenly expanded to 90 hours - or about 4 days. Please note that if left alone, the host would have downloaded more work in a few hours.</blockquote> I think it must have been a problem with that particular WU as once it finished Boinc went out of panic mode and downloaded Wu's for all projects about 16 WU's in total. This included another 2 LHC WU's which are due to be returned on 16/10 about 12 hours before the last WU that went into panic mode was due to be returned. These LHC WU's did not go into panic mode so it must have been a one off situation.
5) Message boards : Number crunching : Reason for not getting work: won't finish in time... (Message 10624) Posted 7 Oct 2005 by Ross Morgan Post: <blockquote>Ross, Post your boinc-project allocations, boinc client uptime %, and boinc client version we can tell you why it's doing that. It took my 10 active hosts about 1.5 months to finally even out to something which resembled predictability. The Boinc client does work correctly and in 99%+ of cases it's right. </blockquote> My project allocaions are:- Einstein 300 (42.86%) Climate Prediction 100 (14.29%)(Currently suspended) Predictor 14.29% 100 (14.29%) LHC@Home 100 (14.29%) Seti 100 (14.29%) Client uptime 63.8777% Boinc Client Version 4.45
6) Message boards : Number crunching : Reason for not getting work: won't finish in time... (Message 10619) Posted 7 Oct 2005 by Ross Morgan Post: I run my P4HT 3.4Ghz for about 12 hours a day average and i have currently got a LHC@home WU being crunched which is not due to be returned until till the 17th of October suspending the work fetch for all projects on my computer because Boinc thinks my PC is overcommitted. This is rubbish as this work unit has only a few hours to go and even if i downloaded work from other projects this LHC@home work unit would be finished later today or tommorrow morning at the latest. There's something well wrong with some of the code in that Boinc work scheduler.
7) Message boards : Number crunching : Resigning due to short deadlines (Message 9897) Posted 3 Sep 2005 by Ross Morgan Post: The deadlines seem to be getting longer again. Just got one for the 12/09/05 today.
8) Message boards : Number crunching : Do we have a 'some-weeks-study-outage'? (Message 8944) Posted 26 Jul 2005 by Ross Morgan Post: > 109609 workunits to crunch and rising. It seems now we have a lack of > new users with fast CPUs to crunch all available workunits in a timely manner. > Don't if that's actually true as all of the few workunits i have completed so far have had at least 3 of the users return them within in a few hours of getting them.

LHC@home