Posts by Angus

1) Message boards : Number crunching : Server Status Page (Message 12576) Posted 30 Jan 2006 by Angus Post: For those that pointed out that LHC@Home does not have SETI@Home characteristics and components shoud be relieved to know that the server status page has been updated ... Where is it??? There doesn't appear to be alink to any server status page from the Home page
2) Message boards : Number crunching : could we get an update about the current progress? (Message 11575) Posted 6 Dec 2005 by Angus Post: Will this project have more work, or is it a dead duck?
3) Message boards : Number crunching : Resigning due to short deadlines (Message 9838) Posted 2 Sep 2005 by Angus Post: <blockquote>This is again a problem with longterm and short term view of things. GasGiant you want your machines not to run out of work, which is understandable, but if we give long deadlines we will have a very long tail of jobs that do not finish. This will then lead to a very long time for a study to finish, and therefore a much longer time between runs where you will have work outages from us. I think we all noticed that the last work outage was on the order of days and not weeks or months as we have seen before. I agree that maybe the current deadline is a bit short, but it is what the statistics say will give us the best turnaround time(well it is actually two standard deviations more to not give a too low deadline), if too many people drop out and stop crunching the statistics will automatically detect this and raise the deadline again. I might incorporate a minimum deadline of say 6-7 days. I do not think that a 5 day deadline is way too low but it is about the smallest reasonable amount. </blockquote> Why can't you have multiple runs going at once, with staggered starts? I'm sure the physicists and engineers can come up with more than one scenario to test at the same time... Then nobody would be in the position of running out of work while they decide what to test next.
4) Message boards : Number crunching : Resigning due to short deadlines (Message 9833) Posted 1 Sep 2005 by Angus Post: That's exactly the problem with short deadlines (those significantly different from the rest of the BOINC projects). As soon as a short deadline WU hits the work queue, it immediately puts BOINC into EDF mode. This causes the other attached projects to start increasing their LTD. When the few short deadlines WUs have finished, the other projects start crunching, and no new short deadline WUs are downloaded. It's a short term hijacking of the user's BOINC PC, forcing the WUs to the front of the queue. I think that it will backfire, both in political and real terms. After a short blip of work activity (maybe a week?), the PCs that are working off the LTD on other projects won't be crunching LHC.
5) Message boards : Number crunching : information should not be that difficult (Message 9783) Posted 31 Aug 2005 by Angus Post: <blockquote>Can somebody explain what is an '...adequate cache of work...'? Adequate to what? What is the interest to have 10 WUs in stock if a project is down for a week. Until now I thought we offer a service. This is less ans less the case, I think. The only important thing with the deadlines is to be able to run several project at a time. I have set my cach on 0.1 day. And I crunch Wu after WU. If a project is down, it raises the percentage allowed to other project. Automatically. I never have a problem. Some people will say that everybody is free to choose the way he wants to work. Yes but in the project limits. Don't forget we offer a service. The goal is not to control the projects. Not for me at least :-p </blockquote> Well, my stated goal since DC was a baby is to gain as many points as possible in whatever project I choose, to advance my standing and my team's standing. I'm not hiding that behind any phony talk about the 'science'. It's pure STATS. If that offends you, too bad.
6) Message boards : Number crunching : Can't get new work (Message 9771) Posted 31 Aug 2005 by Angus Post: <blockquote><blockquote>Cache size is not the same as connection type or interval.</blockquote> At the moment though, it is the same. The cache size is the time between connections. Also, LTD numbers seem to get more accurate (and less inflated) with a lower cache. If you have a constant connection anyway, try a 0.5 day connect time (or even 0.25).</blockquote> NO - I will NOT play that game with cache size. It defeats the whole concept of a cache.
7) Message boards : Number crunching : Resigning due to short deadlines (Message 9770) Posted 31 Aug 2005 by Angus Post: <blockquote><blockquote><blockquote>Why not just let the BOINC Manager do what it was designed to do? If it goes to EDF mode, so what? The resulting LTD for LHC will mean that the other projects get their fair share of CPU time later, and it should all even out in the end. What's the obsession with micro-managing the manager?</blockquote> Because the "Manager" is not managing to the user's set preferences. And Users fuel the projects. And STATS fuel the users, notwithstanding the few percent who really think we're all in it for the science - they are in denial.</blockquote> According to your own post in this thread, Angus, you've quit the project. What difference does it make to you now? </blockquote> Just because I'm not currently crunching WUs here, does not mean I am not monitoring the project. If/when the WU deadlines improve to reasonable settings, I will resume crunching.
8) Message boards : Number crunching : Resigning due to short deadlines (Message 9768) Posted 31 Aug 2005 by Angus Post: <blockquote>Why not just let the BOINC Manager do what it was designed to do? If it goes to EDF mode, so what? The resulting LTD for LHC will mean that the other projects get their fair share of CPU time later, and it should all even out in the end. What's the obsession with micro-managing the manager?</blockquote> Because the "Manager" is not managing to the user's set preferences. And Users fuel the projects. And STATS fuel the users, notwithstanding the few percent who really think we're all in it for the science - they are in denial.
9) Message boards : Number crunching : information should not be that difficult (Message 9759) Posted 31 Aug 2005 by Angus Post: <blockquote><blockquote> For some reason, they have gone to an insanely short deadline. People complained bitterly at Einstein because of this.... I know I would never stay at Einstein with the short deadlines, because I want a large cache - more than a couple of days. </blockquote> This are perfect example of statements made by WU collectors. I don't understand why you need so many results on your puter when they are stored on server. And when they run dry on one project the others will get more attention. That is why BOINC was made. Tony</blockquote> I take offense at your characterizing me as a "WU Collector" for simply wanting to have an adequate cache of work on hand to weather outages, whether they are due to lack of WU, or some other problem like web sites down, databases down, connectivity problems, etc. I don't buy into the statement that BOINC is designed so you can run other projects when one is dry. It is merely a crutch for projects who want to play in the big leagues of DC, but can't generate enough work to satisfy demand. I will agree that it was designed as a platform that projects could utilize without creating their own from the ground up, and that it makes it possible (but NOT MANDATORY)to run multiple projects. I pick the projects I want to run, and the fact that it's under the BOINC umbrella is immmaterial to me. I don't need BOINC deciding where my cycles go. If one project's deadlines are out-of-whack with the others it causes issues with all the projects that you might be attached to, running BOINC into EDF mode immediately, then the LTD dance begins...
10) Message boards : Number crunching : information should not be that difficult (Message 9744) Posted 31 Aug 2005 by Angus Post: <blockquote>we are not getting work. while it is obviously there. lowerring the network connection helps sometimes. why? you have your own set of boinc rules? maybe then you could make this public? </blockquote> For some reason, they have gone to an insanely short deadline. People complained bitterly at Einstein because of this.... I know I would never stay at Einstein with the short deadlines, because I want a large cache - more than a couple of days. I have now done a project reset on all my boxes running LHC. You all can have the WUs in 5 days or so to crunch. Good Luck.
11) Message boards : Number crunching : Can't get new work (Message 9740) Posted 31 Aug 2005 by Angus Post: <blockquote><blockquote><blockquote><blockquote> What gives? Why won't my machine download more LHC units? </blockquote> LHC is now issuing units with a deadline of around 5 days. With your cache set at six days the unit will have expired before your machine uploads the result to LHC. Reduce your cache size and you should get some work when your current queue reduces. </blockquote> Cache size is not the same as connection type or interval. Even though my cache size is 7 to 10 days, the PCs are always connected and should be uploading and downloading whenever a WU is finished. This seems to be a problem in many of the recent BOINC verions. </blockquote> But if your cache size is set to 10 days you can't finish the last workunit you downloaded within the deadline limits. You just download too many workunits at the same time, the last workunits will be past the deadline by the time you want to process them. May I ask you why you take such a big cache size when you are constantly connected because I can't really see the advantage.</blockquote> To be able to keep crunching when work disappears for days... the same reason caching add-ons were written for SETI-1. However, BOINC didn't implement caching properly - the cache size and connect interval need to be separate parameters.
12) Message boards : Number crunching : Can't get new work (Message 9738) Posted 30 Aug 2005 by Angus Post: <blockquote><blockquote> What gives? Why won't my machine download more LHC units? </blockquote> LHC is now issuing units with a deadline of around 5 days. With your cache set at six days the unit will have expired before your machine uploads the result to LHC. Reduce your cache size and you should get some work when your current queue reduces. </blockquote> Cache size is not the same as connection type or interval. Even though my cache size is 7 to 10 days, the PCs are always connected and should be uploading and downloading whenever a WU is finished. This seems to be a problem in many of the recent BOINC verions.
13) Message boards : Number crunching : Even shorter deadlines? (Message 9560) Posted 22 Aug 2005 by Angus Post: It's the problem that's been beat to death over on the SETI boards - The "Cache Size" and the "Connect Every xx Days" need to be separate parameters. That will let people size their cache as desired and still be able to connect and upload/download/report on a timely basis.
14) Message boards : Number crunching : only 10 k WU's left (Message 9480) Posted 19 Aug 2005 by Angus Post: <blockquote>....more work will be submitted soon. Ben Segal / LHC@home</blockquote> Can you define "soon" ? Are we talking hours, days, or weeks? If too long, I'll add another project to pick up after my queue runs dry, otherwise if it's only going be a day or so I'll ride it out with my cache.
15) Message boards : Number crunching : Is there actually going to be some work? (Message 3521) Posted 11 Oct 2004 by Angus Post: There's something fundamentally wrong with projects wanting to use the vast power of Distributed Computing, going to all the trouble to produce and release a client (hopefully a stable one) and then only having enough work for only a very small fraction of the people that would want to run their project. Unless I'm missing something, it appears LHC could operate with only a couple of hundred CPUs. I'm not sure how many WUs are released at once, but 10,000 of the current (last) type we saw could easily be crunched by 150 PCs in a day. Give them 2 or 3 days to do the triple-checks and by then the physicists should be ready again. DC is for projects that can't be accomplished otherwise - it sounds like this one could have been accomplished without a worldwide rollout, and the resulting negative publicity after people switch to it, only to find that it's idle a huge majority of the time.
16) Message boards : Number crunching : Is there actually going to be some work? (Message 3486) Posted 11 Oct 2004 by Angus Post: > > > But what if you dont WANT to join a bunch or projects? The only ones > Im > > > interested in with regards to BOINC are Predictor@Home [which Ive > pretty > > much > > > given up on] and this one [which has been down more than up since I > > joined]. > > > > You should not give up on Predictor... rather, give up on Dell since > that is > > the biggest reason why they aren't back up yet. Dell has taken over a > month > > to get their new server shipped. (and I don't think it is out of the > door > > yet) I believe their upgrade to BOINC v4 and new science applications > are > > nearly complete but they don't want to roll it out until they have new > > hardware. > > > > -------------------------------------- > > A member of <a> href="http://lhcathome.cern.ch/team_display.php?teamid=10">The > > Knights Who Say Ni![/url] > > My BOINC stats site > > Forgive me if Im a little suspicious about that whole thing. Its been over a > month and they have no idea when they are getting their server? Im no DELL > lover by any stretch but this whole waiting for DELL story makes me wonder. > Its not like they are some guy sitting at home waiting for his DELL laptop. > This is a large institution putting great resources behind their purchase. > Most of the people here could have built the server for them by now.... > > Spectre > > > Spectre > If LHC is the ONLY project within BOINC that interests you (or me, in this case) then it's all the more harder to deal with this constant lack of work. I wasn't able to get enough work for 2 boxes to crunch unattended over the weekend. When they run out of work, I have to then manually start another DC application to keep the boxes busy until LHC decides to grace us with a small allotment of work again. Which, if you're following this, means I have to monitor the LHC app or this site to see if there IS work again, and then stop the OTHER application so LHC can continue unimpeded. Hard core DC'rs don't like to have boxes sitting idle - ever. And DC projects NEED the hardcore crunchers, not the casual hobbyist who runs the client for a day or two and then gets bored and goes on to whatever else strikes their fancy. If the boxes can't be kept busy, we'll go on to something else. As for the Dell issue, I have a whole computer room full of Dell servers, both tower models and multiple racks full of various model servers. I have NEVER had to wait that long for a server.
17) Message boards : Number crunching : How long to wait before work is available? (Message 3208) Posted 5 Oct 2004 by Angus Post: > Change to BOINC 4.09. It's latest public release. > Time sharing is much better than in 4.05 > How? I can't find a DL for it on the BOINC site...
18) Message boards : Number crunching : How long to wait before work is available? (Message 3205) Posted 4 Oct 2004 by Angus Post: > > I would start another BOINC project, but the BOINC timeslicing thing is > crap. > > What version of BOINC are you running? What's your OS? Are you connected to > the internet 100% of the time? > > I've found that the time slicing is way off when a project is first added but, > after about a week, the %'s fall right in place. I do not connect to more > that 2 projects on any on host though. > > I am currently testing BOINC 4.11 alpha. > BOINC Version 4.05 No matter how you tweak the sharing %, it will stop work on one project to work on the other for a while. I want it to work on one project 100% of the time, UNLESS it's out of work or otherwise indisposed, then drop to a backup project.
19) Message boards : Number crunching : How long to wait before work is available? (Message 3197) Posted 4 Oct 2004 by Angus Post: If the server status is "Up, but no work" what is the typical wait before work is available again? Is this an hour or so, or days? I'm new to this project, so I have no feel for how often this happens or how long the work outages last. I would start another BOINC project, but the BOINC timeslicing thing is crap. It would be nicer to be able to designate priority project that will crunch unless it's out of work, then a fallback project kicks in.

LHC@home