Message boards :
Number crunching :
Multi Runs of BOINC ?
Message board moderation
Author | Message |
---|---|
KR Jones Send message Joined: 25 Dec 03 Posts: 13 Credit: 155,581 RAC: 0 |
I used to be able to set 10 command line SETI's at once in the background of this system, yes, 10, and still had control of the system and such. The new BOINC does not seem to be processing as fast as that array. I attempted to setup a boinc_gui2 and all hell broke out and I lost some data including a climateprediction.net item that was 60% complete! Had to merge hosts to fix the results of that disaster. Any ideas? I know this system is able to crank out a better rate than the credit I have been given thus far. My last setup in SETI 3.08 generated almost 4,000 work units in almost 6 months of run time. |
Keck_Komputers Send message Joined: 4 Jul 99 Posts: 1575 Credit: 4,152,111 RAC: 1 |
Normally you get the best output by running one process per logical CPU. So BOINC is set up to run that way automatically. BOINC makes certain registry enteries so only one copy can run on a windows system. BOINC WIKI BOINCing since 2002/12/8 |
Toby Send message Joined: 26 Oct 00 Posts: 1005 Credit: 6,366,949 RAC: 0 |
What in the name of all that is holy prompted you to run 10 instances of seti?? Unless of course you have 10 CPUs... The more processes you have running, the more time you are going to lose to context switching between them. Sure you may do 10 work units at a time but they will all take 11 or 12 times as long. As JKeck said, one process per CPU will give you the best performance. BOINC takes care of this automatically if you have your preferences set up right. And actually it isn't anything as fancy as a registry entry - just look in the BOINC directory for the "lockfile" :) A member of The Knights Who Say NI! For rankings, history graphs and more, check out: My BOINC stats site |
KR Jones Send message Joined: 25 Dec 03 Posts: 13 Credit: 155,581 RAC: 0 |
> What in the name of all that is holy prompted you to run 10 instances of > seti?? Unless of course you have 10 CPUs... The more processes you have > running, the more time you are going to lose to context switching between > them. Sure you may do 10 work units at a time but they will all take 11 or > 12 times as long. As JKeck said, one process per CPU will give you the best > performance. BOINC takes care of this automatically if you have your > preferences set up right. And actually it isn't anything as fancy as a > registry entry - just look in the BOINC directory for the "lockfile" :) > It wasn't 10 instances of BOINC , it was the X86 Command line V.3.08. Runs great acually and the system was cranking out 15-20 Work units per day. So I gather the architecture of the way this BOINC calculates and the older CMD LINE is entirely different. It still seems to be slower than the previous CMD LINE 3.08 at leat on this system as I can only turn out 12 maximum per day it seems. |
KR Jones Send message Joined: 25 Dec 03 Posts: 13 Credit: 155,581 RAC: 0 |
> Normally you get the best output by running one process per logical CPU. So > BOINC is set up to run that way automatically. BOINC makes certain registry > enteries so only one copy can run on a windows system. > Is there A DOS Style Command line version of this BOINC capability yet? I think that is the question I have really, because 10 instances of the old GUI version of SETI would not have worked on a system, but the CMD line 3.08 did. |
Divide Overflow Send message Joined: 3 Apr 99 Posts: 365 Credit: 131,684 RAC: 0 |
You can run BOINC as a Windows service. I have no idea if you can run multiple sessions of this service, but I doubt it. I think that you need to shed some preconceptions about BOINC that you are carrying over from Seti@Home classic. The program attempts to make maximum use of the processing resources available. If you have multiple CPU's, BOINC applications will use them if you set your configuration to do so. The WU's here are not the same as the Classic work units, either. So a direct comparison of WU's crunched on BOINC to Classic is useless. For some excellent BOINC documentation, inclucding Seti@Home and other BOINC projects, check Paul's site here. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
> Is there A DOS Style Command line version of this BOINC capability yet? > Check in the directory you put BOINC in for a file called boinc_cli.exe It's only available in the non-alpha versions of BOINC, meaning 4.13, 4.14, 4.15 and 4.16 do have it. If you're running an Alpha version (4.5x, 4.6x), it's getting a different story totally, as boinc.exe will always run (be it as a service, or started by you), while you control in through the boinc manager (seperate program). Just run boinc_cli /? for options. |
SURVEYOR Send message Joined: 19 Oct 02 Posts: 375 Credit: 608,422 RAC: 0 |
4.14, 4.15, 4.16 still in testing. 4.16 should be releast soon Fred BOINC Alpha, BOINC Beta, LHC Alpha, Einstein Alpha |
Marco Niese Send message Joined: 21 May 99 Posts: 11 Credit: 4,238 RAC: 0 |
I know this isn't the wishlist, but I'd rather have some switches in the gui (for use in the icon), something like: Target "C:|Program Files|BOINC|Boinc_Gui.exe" -run_always -hide - Marco <a href="http://setiweb.ssl.berkeley.edu/team_display.php?teamid=36971">Team #LuckyStar</a> <img src="http://www.boincstats.com/stats/banner.php?cpid=e3c78d10c4cf326b67e1210e1db0ce55"> |
bjacke Send message Joined: 14 Apr 02 Posts: 346 Credit: 13,761 RAC: 0 |
What is faster the commandline or the intervace? WARR - Wissenschaftliche Arbeitsgemeinschaft für Raketentechnik und Raumfahrt (WARR - scientific working group for rocket technology and space travel) |
Divide Overflow Send message Joined: 3 Apr 99 Posts: 365 Credit: 131,684 RAC: 0 |
From what I've seen personally, there is little or no difference in WU completion times from the "command line" service and the GUI, when the GUI is run without screensaver graphics and left running minimized while on the Projects tab. |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
Mr. KR Jones, I looked at your computers via clicking on your link here in the forums, then on view computers. From my short study it appears your 3.4Ghz Pentium 4 is completing WUs in an average of 11,400 seconds. And it is hyperthreaded so 2 WUs are being computed at the same time. Therefore, your system should output 86400 / 11400 * 2 WUs per day, or about 14-15 which is what you said you were getting before (you said 15-20, WU times vary by about +/- 15%) So...is there a question? |
KR Jones Send message Joined: 25 Dec 03 Posts: 13 Credit: 155,581 RAC: 0 |
> Mr. KR Jones, > > I looked at your computers via clicking on your link here in the forums, then > on view computers. > > From my short study it appears your 3.4Ghz Pentium 4 is completing WUs in an > average of 11,400 seconds. And it is hyperthreaded so 2 WUs are being > computed at the same time. > > Therefore, your system should output 86400 / 11400 * 2 WUs per day, or about > 14-15 which is what you said you were getting before (you said 15-20, WU times > vary by about +/- 15%) > > So...is there a question? > > Actually the BOINC GUI shows the resources split between SETI and ClimatePrediction.net. And the "work" tab shows one work unit in analisys. When I eliminate Climate Prediction the 2 w/u's are in progress, but I still feel the CMDLINE 3.08 seemed faster. I did almost 4000 work units in about 6 months. I wish I had actual numbers so i could calculate the exact units per day. |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
Eric Korpela wrote the seti science app, for both classic and BOINC. Only small modifications were made to communicate with BOINC servers, etc. The WU calculation code only had changes made for slight structure changes in WU packages. |
Dave Mickey Send message Joined: 19 Oct 99 Posts: 178 Credit: 11,122,965 RAC: 0 |
I have to agree that the Classic CLI 3.08 and the current SAH/BOINC application are very close in runtime on the 3 machines on which I have run both versions. Certainly less than 10% variance, and more likely inside 5%. It was close enough that it never seemed any different, in human terms. On each, I get 11-12, 18-20, and 36-38 hour runtimes on 500, 333C, and 233 MHz machines, rather predictably, other than the occasional oddball short WU. Dave |
7822531 Send message Joined: 3 Apr 99 Posts: 820 Credit: 692 RAC: 0 |
Classic's runtime averaged around 11 hrs with graphics and 8 hrs without. SETI/BOINC runs about 7 hrs per WU. |
KR Jones Send message Joined: 25 Dec 03 Posts: 13 Credit: 155,581 RAC: 0 |
> Eric Korpela wrote the seti science app, for both classic and BOINC. Only > small modifications were made to communicate with BOINC servers, etc. The WU > calculation code only had changes made for slight structure changes in WU > packages. > > > Well I guess I'm just not elloquent enough to get the right question out. It may be that the calculation time is nearly equivalent from one version to another, however, I can confirm that between mid-May and mid-December, my PC, while running 10 instances of 3.08 CMDLINE, generated 3500+ work units. Thus averaging 19.4 per day. BOINC_GUI is doing one every 2 hours, steady, for a 10-12 average per day. ClimatePrediction.net also runs BOINC on this machine. |
KWSN - MajorKong Send message Joined: 5 Jan 00 Posts: 2892 Credit: 1,499,890 RAC: 0 |
> > Eric Korpela wrote the seti science app, for both classic and BOINC. > Only > > small modifications were made to communicate with BOINC servers, etc. > The WU > > calculation code only had changes made for slight structure changes in > WU > > packages. > > > > > > > > Well I guess I'm just not elloquent enough to get the right question out. > > It may be that the calculation time is nearly equivalent from one version to > another, however, I can confirm that between mid-May and mid-December, my PC, > while running 10 instances of 3.08 CMDLINE, generated 3500+ work units. > > Thus averaging 19.4 per day. > > BOINC_GUI is doing one every 2 hours, steady, for a 10-12 average per day. > > ClimatePrediction.net also runs BOINC on this machine. > Running CPDN on that machine will take some cycles away from seti. if the resources are split 50%/50%, Seti will only get about 1/2 the cycles. If, running full out on nothing but seti-classic, your machine did around 20 a day, and if it does around 10 a day with CPDN also running, and you are splitting the time evenly, then the seti/boinc speed and the seti-classic speed are equivalent. That said, running more instances of seti-classic than you have CPUs is not really a good idea. The context switches between the instances will eat away at your available cpu time. https://youtu.be/iY57ErBkFFE #Texit Don't blame me, I voted for Johnson(L) in 2016. Truth is dangerous... especially when it challenges those in power. |
KR Jones Send message Joined: 25 Dec 03 Posts: 13 Credit: 155,581 RAC: 0 |
All I can say is when I had one instance of the commandline 3.08 running on this machine, I did not get the same number of work units per day out as I did when I ran 10. I averaged what I am averaging now 10-12. When I made multiple folders, changed the program name (ex.: seti2.exe,etc) and enabled them at startup for 10 instances, I almost doubled my output. Maybe it has to do with the Xeon Architecture. I believe a good percentage of the top 25 machines are Xeon. My Extreme edition is a "gene altered" bastard cousin of the Xeon. New thought : Ok, I see now where BOINC works, you're correct, if i skipped CPDN, I would equal my old output. Maybe what I was attempting to get across before was that the mult-instance CMDLINE was able to swing what BOINC does now, except BOINC, I guess, is more efficient at it. |
Hans Dorn Send message Joined: 3 Apr 99 Posts: 2262 Credit: 26,448,570 RAC: 0 |
> All I can say is when I had one instance of the commandline 3.08 running on > this machine, I did not get the same number of work units per day out as I did > when I ran 10. OK, since your cpu supports hyperthreading, you didn't make full use of it with only one instance of seti running. The optimal number of instances might have been 2 (one for each virtual cpu) or 3 (to keep crunching at full speed while downloading new work). Regards Hans |
©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.