Thanks very much for wasting a shit load of my time and my CPU time. In future when you are going to make major changes which will inturn stop a machine from running and make the hours of work it has done useless you might want to consider first sending out a few emails to registered users.
Many people cannot afford to spend the time checking the website and the forums for changes every day and when a change like this is happening send an email to registered users. You have stuffed me around and this pisses me off to no end. In future think about these things first or this fairly major contributor will DPAD no longer
|Something to keep in mind is that you _voluntarily_ contribute your own time and resources to this project.|
I can understand being frustrated - I lost, at the very least 5,000,000 Mpts of production. Oh well... Intermittent problems are a fact of life and that’s something everyone should accept. But the recent problems associated with the new client are no grounds for scolding the sole developer of a fairly solid distributed computing project.
Stephen has done an excellent job with addressing the issues that came up with the latest client release. If you find a bug and mention it on the forum he typically responds within a few hours (if not much sooner). Count your blessings and realize that it is the rare exception to distributed computing projects that the developer consistently interacts with a project’s user base.
Remember that Stephen is a one man team holding down a fulltime job while finishing his masters in math. He built this thing in his free time as an aide to his work on the neutrino factory design at RAL.
If anything, Mr. Brooks is deserving of a little more support from all of us.
|5 million mpt!! Holy cow you can produce that much in 1 day if I were to lose that much my score would go down to 0.|
quote:I don't usually make those sort of changes intentionally. What exactly happened to yours? Did it download the ...B lattices and then complain "v4.41 isn't installed" and quit? Maybe that shouldn't have been put in as a fatal error since you could still have used 4.4 on "SolenoidsTo15cm" successfully. Anyway, the reason I produced the new scripts was that the old ones had bugs in them, so really your work on those wouldn't have been that useful had you been running at the time or not.
quote:Well either the clients well all bring up a message box saying "v4.41 is out" when they try to send, or you will get another hint such as them saying the new lattices aren't compatible with your version.
If you want to preserve the rest of your CPU cycles with 100% reliability, how about running Muon1 on "L"ow priority and having another task like SETI running at background priority so it mops up all the rest and sends the work to your preferred team?
Well I actually lost over 48 million mpts in a day, but what I am referring to is my loss in production that followed that. The changes made caused 50 or so of my dpad machines to stop yielding results. I monitor them on a daily basis, and what I saw was the program running at 99-100% utilization consistently on all of them.
Unfortunately it took me some time to realize that none of the results.txt files had been updated since 05/13 (the day I lost all of my results). I probably would have caught this sooner, but was assuming all along that my points were being dropped or not counted on a daily basis because of the other stats issues. I hadn't even considered that results weren't even being produced on my end.
quote:I don't think so... 48 million Mpts is roughly your entire production to date, and your file for v4.34 is still intact, as are your stats.
Maybe also a better way of monitoring your computers would be by results.dat length, if you can do that?
|It was when the rawstats file was not updating properly... A score of a hundred thousand or so was shown (some 48+ million less than was showing before). I'm not saying I lost that production - Just that the stats showed it as not being there for a day or so. The 5 million I essentially did lose from "non working but busy" clients over a period of about a week.|
I suppose I could make something to report results.dat sizes to me over some interval, but I'm not quite sure how that would work if the client is still downloading/merging sample results.
quote:Again you might want to consider first sending out a few emails to registered users. I do not have enough time to check Distributed Folding (Rank 19th), F@H (Rank 4th) and Muon Forums every day. I don't care that a few results were going to be invalid, but if you learn to manage information excahnge about these changes better, ie emails then I would have been able to stop clients from doing those optimisations which I had set manually and set them to 15cm only until the new client was available
That does jack when you are running all your machines as a service!
SETI, I like real science not Alien myths (awaits flames). Check out DF and F@H, I have been No.1 contributors in both and still am currently ranked highly in both, but I have my machines dedicated to projects. I have 30 machines which i personally own and operate for these projects and my Muon machines are in a location with only remote access so I cannot have F@H or DF running as they don't fair so well running offline. I thank you for your suggestions but Stephen, these things can be more easily fixed at your end with a simple email
Totally with you on that
Who needs a computer to hear an alien? There are plenty of females here on Earth.
|As said before - if stats are really that important, don't do this project.|
Maybe something similar should be stated on the main page.
Funny I don't see that I said stats were important, infact I never mentioned anything ti do with lost stats, but I dislike the wasted money and time I spend on idle clients. I don't give a shit how big my stats are points wise, I care more about the number of results and the total time I have been able to contribute and when several machines sit idle not because of my side and this could have been managed better I get annoyed to say the least.
Please read the thread before making comments
|Who said that it was wasted effort? One of the reason why PhaseRotB optimised so fast was because that of the PhaseRotA results that could be manually seeded into the new client. ChicaneLinac is another matter and even with the older slower client it wasn't moving very fast at all.|
|Yes, I see how it could be annoying - I'm not entirely sure when you would want me to send an e-mail out? At the same time I release a version on the web, for instance?|
Mostly this problem was caused because although I tested v4.41 running the "B" optimisations on its own, I forget to test what would happen to an already existing v4.4 client when the changeover occurred. This has tipped up 2 bugs: one was that when my "don't mind if relevant lattice files aren't there when sending" fix went in, I didn't fix another place it looked in; the other was this business with the "lattice requires higher version" error being fatal, when really it should have been a warning and the program should carry on with any compatible lattices still active.
So really, in an ideal world I wouldn't have to send out e-mails except when I plan to upgrade in a way that really does require immediate user intervention. The v4.4 --> 4.41 upgrade shouldn't have been quite that drastic, though I did want rid of the old clients and bad-ish optimisations ASAP.
Another idea: what if http://stephenbrooks.org/muon1/versions.html was available as an RSS feed? Since I update that with every new version you would be able to use a news tool to subscribe to that.
Well, since you bragged about your rank on other projects, what was I supposed to think?
|AySz88 You were supposed to think that here is a guy that is dedicated and experience in many projects, especially ones that deal with real research, and also obviously has had to deal with managing many machines and obviously quite sucessfully and is not a fly by night whinger|