Message boards :
Number crunching :
This is how the perfect WU looks like...
Message board moderation
Author | Message |
---|---|
. Send message Joined: 3 Apr 99 Posts: 410 Credit: 16,559 RAC: 0 |
Just browsing my credit and saw this: [url=http://setiweb.ssl.berkeley.edu/workunit.php?wuid=6267029] http://setiweb.ssl.berkeley.edu/workunit.php?wuid=6267029 <a> This is a perfect example on how a WU can be processed, when it's at it's best! :-) |
Toby Send message Joined: 26 Oct 00 Posts: 1005 Credit: 6,366,949 RAC: 0 |
It is indeed a beauty as far as timing and validation. Unfortunately that non hyperthreading CPU got screwed on the credit by the 2 hyperthreading ones :) Ah, the joys of the benchmark system! A member of The Knights Who Say NI! For rankings, history graphs and more, check out: My BOINC stats site |
bjacke Send message Joined: 14 Apr 02 Posts: 346 Credit: 13,761 RAC: 0 |
> It is indeed a beauty as far as timing and validation. Unfortunately that non > hyperthreading CPU got screwed on the credit by the 2 hyperthreading ones :) > Ah, the joys of the benchmark system! > That's realy true ;-). WARR - Wissenschaftliche Arbeitsgemeinschaft für Raketentechnik und Raumfahrt (WARR - scientific working group for rocket technology and space travel) |
ghstwolf Send message Joined: 14 Oct 04 Posts: 322 Credit: 55,806 RAC: 0 |
> It is indeed a beauty as far as timing and validation. Unfortunately that non > hyperthreading CPU got screwed on the credit by the 2 hyperthreading ones :) > Ah, the joys of the benchmark system! > It happens. There is no perfect system, but I'll give credit where it is due, this one is pretty good. Eventually, they may seperate processor types (one of the many someday upgrades). The wishlist is pretty long, and in spite of the cool factor of credits, I'd rather see the effort go to forwarding the project (yeah I know this is a part of it). Still looking for something profound or inspirational to place here. |
UBT - Timbo Send message Joined: 3 Apr 99 Posts: 157 Credit: 10,720,947 RAC: 362 |
> > It is indeed a beauty as far as timing and validation. Unfortunately > that non > > hyperthreading CPU got screwed on the credit by the 2 hyperthreading ones > :) > > Ah, the joys of the benchmark system! > > > > It happens. There is no perfect system, but I'll give credit where it is due, > this one is pretty good. Eventually, they may seperate processor types (one > of the many someday upgrades). The wishlist is pretty long, and in spite of > the cool factor of credits, I'd rather see the effort go to forwarding the > project (yeah I know this is a part of it). > Hang on a second...! If the SETI host database system knows the type and speed of the host processor (coz it's in our "account info"), what is to stop BOINC only downloading specific WU's to a group of hosts that all have the exact same type and speed of host? That way, everyone who has a particular speed of non-HT Pentium 4 will get the same WU, while those with say a HT Pentium 4, would get WU's froma different batch. Likewise, the type of OS can also be interrogated to ensure that Linux crunched WU's are not compared/credited against Windows crunched WU's. This might then allow the benchmark system to work better. Of course there is a downside - some OS (or hardware) maybe very rare and hence these hosts might not get much credit as there would be few other systems with identical OS/CPU. But as soon as SETI Classic stops, surely there'll be enough hosts with identical spec's to make this only a small issue. Just my 2p's worth. Timbo regards, Tim Founder, UK BOINC Team Join us @ UK BOINC Team: http://www.ukboincteam.org.uk/newforum |
Hans Dorn Send message Joined: 3 Apr 99 Posts: 2262 Credit: 26,448,570 RAC: 0 |
> Hang on a second...! > > If the SETI host database system knows the type and speed of the host > processor (coz it's in our "account info"), what is to stop BOINC only > downloading specific WU's to a group of hosts that all have the exact same > type and speed of host? > > That way, everyone who has a particular speed of non-HT Pentium 4 will get the > same WU, while those with say a HT Pentium 4, would get WU's froma different > batch. > > Likewise, the type of OS can also be interrogated to ensure that Linux > crunched WU's are not compared/credited against Windows crunched WU's. > > > This might then allow the benchmark system to work better. > > Of course there is a downside - some OS (or hardware) maybe very rare and > hence these hosts might not get much credit as there would be few other > systems with identical OS/CPU. But as soon as SETI Classic stops, surely > there'll be enough hosts with identical spec's to make this only a small > issue. > > Just my 2p's worth. > > Timbo > Hi, I don't think this would be fair. If you validate linux HT machines against each other, they'll only claim (and get) 1/4 of the credit of, say, an athlon. They'll be off to cpdn or other projects in a breeze. Maybe the benchmark system should be fixed, so that all hosts claim similar credit for the same WU. Regards Hans |
Mike Send message Joined: 17 Feb 01 Posts: 34255 Credit: 79,922,639 RAC: 80 |
Hi What does fair means??? I´ve many of results with 40 to 50 credits claimed and getting granted about 20 credits because two linux machines returned the other two results. By the way it doesnt matter. fun isn´t true and what is true isn´t fun. Mike With each crime and every kindness we birth our future. |
Hans Dorn Send message Joined: 3 Apr 99 Posts: 2262 Credit: 26,448,570 RAC: 0 |
> Hi > > What does fair means??? > According to http://boinc.berkeley.edu/credit.php , 1 credit should be worth 864 seconds of cpu time on a 1GFlop machine, or 864 GFlops. (864,000,000,000 floating point operations). I hope I put the decimal point in the right place here :o) You'd have to count (or estimate) the floating point ops required for a WU to get a fair number. Since _any_ kind of change would mean that some people get more credit, and some people get less, I don't think everybody will agree when changes are made :o) Regards Hans |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
I suggested back in the beta that the CPU type, speed, and OS be considered as a means of classification and got told that this would create a very large and unmanageable data set. Not having access to the data stored in the host data table I do not know if that is true or not. At any rate, this is one area that will be revisited once we take care of the trivial problems of the cross-platform GUI and migration of all participants off of the SETI@Home CLassic project. |
karthwyne Send message Joined: 24 May 99 Posts: 218 Credit: 5,750,702 RAC: 0 |
> I suggested back in the beta that the CPU type, speed, and OS be considered as > a means of classification and got told that this would create a very large and > unmanageable data set. Not having access to the data stored in the host data > table I do not know if that is true or not. > > At any rate, this is one area that will be revisited once we take care of the > trivial problems of the cross-platform GUI and migration of all participants > off of the SETI@Home CLassic project. > personally, do don't really see how this model is that different from the 1 WU = 1 credit. yes it isn't perfect, but the reality was, and still is, a faster machine will get more credit. if my celeron 400 takes 2 days and gets 20 credit, and my HT p4 3.0 does it in 2 hours and get 20 credit, what's the difference. so some work units take longer, and some are really short (same thing, my celeron might take 3 minutes on a WU that the p4 does in 0 seconds) i'm not complaining, mind, we are here for the science/stress testing or some reason - if you are here soley to get credit and fight for position, i just don't understand you or your motives... S@h Berkeley's Staff Friends Club |
B-Roy Send message Joined: 4 May 03 Posts: 220 Credit: 260,955 RAC: 1 |
one thing that i want to mention is that i had no short wu since ages. in classic nearly 10% were short wus (some minutes). i think that the credit system is ok, however i would like to have a page showing some of the project progress: - total wus chrunched by boinc - most interesting signal(s) - certificates - something like a progress bar of crunching (i do not know if this makes any sense, but i would like to know if we crunch faster than data is received in order to see when the break-even point is reached) |
JAF Send message Joined: 9 Aug 00 Posts: 289 Credit: 168,721 RAC: 0 |
I think it's a good idea to crunch WU's with a wide spectrum of computers; slow and fast, different architectures and operating systems. Of course that means a lot of tweaking to keep some semblance of "fair" credit, but it also helps ensure the system (Boinc Seti) as a whole, is functioning world wide, rich poor. <img src='http://www.boincsynergy.com/images/stats/comb-912.jpg'> |
UBT - Timbo Send message Joined: 3 Apr 99 Posts: 157 Credit: 10,720,947 RAC: 362 |
> > Maybe the benchmark system should be fixed, so that all hosts claim similar > credit for the same WU. Hi Hans, My suggestion was that all hosts running similar OS and having similar hardware (CPU, Memory) all get *sent* the same WU and hence will return very similar results, which can then be more easily validated (and every-one gets a similar credit). Least it'll stop the current problem of Windows hosts being credited less than the same hardware running *Nix when crunching the "same" WU. Timbo regards, Tim Founder, UK BOINC Team Join us @ UK BOINC Team: http://www.ukboincteam.org.uk/newforum |
UBT - Timbo Send message Joined: 3 Apr 99 Posts: 157 Credit: 10,720,947 RAC: 362 |
> I suggested back in the beta that the CPU type, speed, and OS be considered as > a means of classification and got told that this would create a very large and > unmanageable data set. Not having access to the data stored in the host data > table I do not know if that is true or not. Hi Paul, And they mean to say that there current database "isn't" large and unmanageable...given what they are already doing - all I'm suggesting is a simple look-up table that assigns certain WU's dependant on a host specification...! Hells teeth - they'll be telling us next that their database is running on a 'pooter using bubble-core memory...! Timbo regards, Tim Founder, UK BOINC Team Join us @ UK BOINC Team: http://www.ukboincteam.org.uk/newforum |
Ned Slider Send message Joined: 12 Oct 01 Posts: 668 Credit: 4,375,315 RAC: 0 |
> > Hi, > > I don't think this would be fair. If you validate linux HT machines against > each other, they'll only claim (and get) 1/4 of the credit of, say, an > athlon. > > They'll be off to cpdn or other projects in a breeze. > > Maybe the benchmark system should be fixed, so that all hosts claim similar > credit for the same WU. > > Regards Hans > Quite right Hans, hence my efforts to produce optimised boinc clients for linux to help reduce the disparity between the two OSes. Now all we need is for all the linux users to actually use the optimised clients that I've made available. We're getting there slowly as I've now logged over 1,000 downloads of my optimised clients :) Ned *** My Guide to Compiling Optimised BOINC and SETI Clients *** *** Download Optimised BOINC and SETI Clients for Linux Here *** |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
> My suggestion was that all hosts running similar OS and having similar > hardware (CPU, Memory) all get *sent* the same WU and hence will return very > similar results, which can then be more easily validated (and every-one gets a > similar credit). The code for this was built into BOINC. The seti project chose not to turn on this setting (the code is there, the flag says, don't use it). LHC was using part of this code because their science app. had large result differences on different CPUs (and perhaps OSes). It is called the "Homgeneous Redundancy" section of the code. With Seti's science app., the results returned by different CPUS are similar enough that they didn't need to turn it on. |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
> personally, do don't really see how this model is that different from the 1 WU > = 1 credit. > yes it isn't perfect, but the reality was, and still is, a faster machine will > get more credit. if my celeron 400 takes 2 days and gets 20 credit, and my HT > p4 3.0 does it in 2 hours and get 20 credit, what's the difference. so some > work units take longer, and some are really short (same thing, my celeron > might take 3 minutes on a WU that the p4 does in 0 seconds) I am not sure why you annswered me with this, but, I was just commenting. My own take on Credit issues is borne out by my tirades in the BOINC FAQ (Credit section). I do look at my numbers, but it is only for amusement, and I don't get hung up on it. I do find it funny that some many people are all about the credit measurement and not about the science. > i'm not complaining, mind, we are here for the science/stress testing or some > reason - if you are here soley to get credit and fight for position, i just > don't understand you or your motives... Actually, I am here as it gets me "out of the house" and I have a chance to "talk" to people. And when I am up for it, I do a little BOINC Documentation on the side to keep me occupied in my copious free time. :) |
Martin Johnson Send message Joined: 9 Jun 01 Posts: 201 Credit: 224,995 RAC: 0 |
B-Roy, I think a bug was introduced in the last version of Classic. If Classic is running when I go on line to get e-mail or browse, it often curtails the current WU. This happened on my old Win 98 m/c, and now on my new Win 2K pro. If I exit Classic, then go on line, then log off and restart Classic, this never happens. I was going to mention it, but thought it pointless if Classic is to be ended soon. |
mikey Send message Joined: 17 Dec 99 Posts: 4215 Credit: 3,474,603 RAC: 0 |
> Just browsing my credit and saw this: > > [url=http://setiweb.ssl.berkeley.edu/workunit.php?wuid=6267029] > http://setiweb.ssl.berkeley.edu/workunit.php?wuid=6267029 <a> > > This is a perfect example on how a WU can be processed, when it's at it's > best! :-) > You want the absolutely perfect unit? Take a look at these stats: Results: Result ID click for details Host Sent Time reported or deadline (future | past) Server state explain Outcome explain Client state explain CPU time (sec) claimed credit granted credit 24363816 370456 21 Dec 2004 9:17:37 UTC 21 Dec 2004 13:07:25 UTC Over Success Done 9,084.25 29.61 29.61 24363817 227195 21 Dec 2004 9:17:38 UTC 21 Dec 2004 17:15:34 UTC Over Success Done 13,796.42 26.91 29.61 24363818 139490 21 Dec 2004 9:17:39 UTC 21 Dec 2004 20:58:07 UTC Over Success Done 8,716.47 39.45 29.61 3 different computers all got the same unit on the 21st of Dec 04, all 3 different computers returned the same unit on the 21st of Dec 04! All 3 returned units were good and credit was issued on 21st Dec 04!!! |
The Gas Giant Send message Joined: 22 Nov 01 Posts: 1904 Credit: 2,646,654 RAC: 0 |
No...I'd say this is the perfect WU. Different machines - but all 'doze. http://setiweb.ssl.berkeley.edu/workunit.php?wuid=6094057 Paul (S@H1 8888) And proud of it! |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.