| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: ud vs boinc, pros and cons hmmm switch or not? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 16
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I had a thought: the fairest way without a controlled environment with 2 IDENTICAL computers is simply to run both agents on the same machine.
UD and BOINC have been observed to split crunching time exactly down the middle with perfect fairness when running simultaneously. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I had a thought: the fairest way without a controlled environment with 2 IDENTICAL computers is simply to run both agents on the same machine. UD and BOINC have been observed to split crunching time exactly down the middle with perfect fairness when running simultaneously. OK, that's a good design for an experiment. It eliminates the somewhat shaky assumption that any 2 machines can be absolutely identical. That aspect of my experiment has always given me some doubts but I think my design sidesteps that issue. The reason I didn't do it your way is because I wasn't sure how UD and BOINC might affect each other's performance when running side by side on the same machine. There were too many potential issues in my mind that I just could not nail down. I did come up with a scheme which, in my mind, eliminates a lot of issues I was unable to get a grip on. I reckon that any dissimilarities between my identical machines will "cancel out" in the second half of the experiment where I intend to test BOINC on the machine that now runs UD and test UD on the machine that now runs BOINC. If the difference in speed between the 2 agents does not change in the second half then I can conclude that neither gained an advantage by running on a slightly faster machine. In the meanwhile, using 2 machines allows me to sample twice as many WUs in the same amount of time compared to running UD and BOINC simultaneously one 1 machine. -- |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges: 
|
I do hope you uncover something new given the effort you're making. I would also recommend looking back thru the several past threads on this subject here (assuming you haven't already) and even in the Chat forum (David Autumns' thread there covered his detailed look at this). At the least though, you will satisfy your own curiousity and it's a great way to learn more about BOINC itself. While you're not the first to look into this in some detail, rest assured that you won't be the last either
---------------------------------------- It would seem there are lots of folks out there from Missouri  Oh, FWIW, someone pointed to the Avg. Points Per Hour of Run Time value as the best way to make comparasions. I want to believe it was one of the admins or community advisors but with the way my memory is these days, it could have been the tooth fairy  By that measure, that value for the time I've been running BOINC has been pretty much double what it was running UD on the same machine. Of course, UD and BOINC have different definitions of run time too so even that is not a perfect comparasion. Also, as UD awards points and BOINC award credits, WCG has to convert BOINC credits into points and that introduces the question of the accuracy of that conversion.---------------------------------------- [Edit 2 times, last edit by keithhenry at Apr 29, 2006 1:08:06 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Keith,
I'm not from Missouri, I only have to be shown once  But until then I am skeptical.Thank you for pointing out that BOINC and UD define run time differently. I foolishly assumed the run times are accurate. I will have to investigate that further. Actually, this is not a lot of work. The computers are doing all the work. All I do is observe what they report. If you are interested in seeing their reports then look at the stats for members BOINC-XP-32 and UD-XP-64. The first has run only BOINC from the start, the latter has run only UD. Note that after close to 20 WUs apiece they are neck in neck. At times BOINC has been ahead by a few minutes while at other times UD has been ahead by a few minutes. Of course, the numbers are meaningless if the run times are not reported accurately. With regard to points or credits.... they have no cash value so they are worthless to me. It would be nice if they were awarded fairly but it's not really an issue I'm concerned about. -- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
... Of course, UD and BOINC have different definitions of run time too so even that is not a perfect comparasion. I think I might have got to the bottom of this one. There are 2 things happening with the run times. First, if you click on BOINC's Messages tab on your computer you will see that it records time in (hh:mm:ss). If you look at your stats on the WCG website you will see that your run times are recorded in decimal hours. In other words, 5:30 from BOINC becomes 5.5 on the website stats page but they are both 5 hours and 30 minutes. Second, the run times on the website for WUs I have completed are about 90 seconds less than the run times I calculate from the WU start and stop times on BOINC's Messages screen. Perhaps there is an error in the code that converts from BOINC's hh:mm:ss format to WCG's decimal format. Anyway, it's not a big error and the Average Run Time Per Unit figure will be off by only 90 seconds for WUs crunched by BOINC. Are the UD run times shown accurately on the WCG website? I can't answer that right now because UD does not seem to keep a record of WU start and stop times. I have started using UDMON in hopes that it will record WU start/stop times. -- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
No, you've missed the point entirely! :-S
UD and BOINC have different definitions of run time. UD uses real time (wall clock time), i.e. the total duration. BOINC uses CPU time. Operating Systems track exactly how many nanoseconds they spend on each process, and BOINC only counts the time that BOINC is working. |
||
|
|
|